| { | |
| "paper_id": "I05-1012", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T07:26:20.326018Z" | |
| }, | |
| "title": "Confirmed Knowledge Acquisition Using Mails Posted to a Mailing List", | |
| "authors": [ | |
| { | |
| "first": "Yasuhiko", | |
| "middle": [], | |
| "last": "Watanabe", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Ryukoku University", | |
| "location": { | |
| "postCode": "520-2194", | |
| "settlement": "Seta, Otsu", | |
| "region": "Shiga", | |
| "country": "Japan" | |
| } | |
| }, | |
| "email": "watanabe@rins.ryukoku.ac.jp" | |
| }, | |
| { | |
| "first": "Ryo", | |
| "middle": [], | |
| "last": "Nishimura", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Ryukoku University", | |
| "location": { | |
| "postCode": "520-2194", | |
| "settlement": "Seta, Otsu", | |
| "region": "Shiga", | |
| "country": "Japan" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Yoshihiro", | |
| "middle": [], | |
| "last": "Okada", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Ryukoku University", | |
| "location": { | |
| "postCode": "520-2194", | |
| "settlement": "Seta, Otsu", | |
| "region": "Shiga", | |
| "country": "Japan" | |
| } | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "In this paper, we first discuss a problem of developing a knowledge base by using natural language documents: wrong information in natural language documents. It is almost inevitable that natural language documents, especially web documents, contain wrong information. As a result, it is important to investigate a method of detecting and correcting wrong information in natural language documents when we develop a knowledge base by using them. In this paper, we report a method of detecting wrong information in mails posted to a mailing list and developing a knowledge base by using these mails. Then, we describe a QA system which can answer how type questions based on the knowledge base and show that question and answer mails posted to a mailing list can be used as a knowledge base for a QA system.", | |
| "pdf_parse": { | |
| "paper_id": "I05-1012", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "In this paper, we first discuss a problem of developing a knowledge base by using natural language documents: wrong information in natural language documents. It is almost inevitable that natural language documents, especially web documents, contain wrong information. As a result, it is important to investigate a method of detecting and correcting wrong information in natural language documents when we develop a knowledge base by using them. In this paper, we report a method of detecting wrong information in mails posted to a mailing list and developing a knowledge base by using these mails. Then, we describe a QA system which can answer how type questions based on the knowledge base and show that question and answer mails posted to a mailing list can be used as a knowledge base for a QA system.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Because of the improvement of NLP, research activities which utilize natural language documents as a knowledge base become popular, such as QA track on TREC [1] and NTCIR [2] . However, these QA systems assumed the user model where the user asks what type questions. On the contrary, there are a few QA systems which assumed the user model where the user asks how type question, in other words, how to do something and how to cope with some problem [3] [4] [7] . There are several difficulties in developing a QA system which answers how type questions, and we focus attention to two problems.", | |
| "cite_spans": [ | |
| { | |
| "start": 157, | |
| "end": 160, | |
| "text": "[1]", | |
| "ref_id": "BIBREF0" | |
| }, | |
| { | |
| "start": 171, | |
| "end": 174, | |
| "text": "[2]", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 449, | |
| "end": 452, | |
| "text": "[3]", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 457, | |
| "end": 460, | |
| "text": "[7]", | |
| "ref_id": "BIBREF5" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "First problem is the difficulty of extracting evidential sentences by which the QA system answers how type questions. It is not difficult to extract evidential sentences by which the QA system answers what type questions. For example, question (Q1) is a what type question and \"Naoko Takahashi, a marathon runner, won the gold medal at the Sydney Olympics\" is a good evidential sentence for answering question (Q1).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "(Q1) Who won the gold medal in women's marathon at the Sydney Olympics? (DA1-1) Naoko Takahashi.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "It is not difficult to extract this evidential sentence from natural language documents by using common content words and phrases because this sentence and question (Q1) have several common content words and phrases. On the contrary, it is difficult to extract evidential sentences for answering how type questions only by using linguistic clues, such as, common content words and phrases. For example, it is difficult to extract evidential sentences for answering how type question (Q2) because there may be only a few common content words and phrases between the evidential sentences and question (Q2).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "(Q2) How can I cure myself of allergy? (DA2-1) You had better live in a wooden floor.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "(O2-1-1) Keep it clean. (O2-1-2) Your room is always dirty. (DA2-2) Drink two spoonfuls of vinegar every day.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "(QR2-2-1) I tried, but, no effect.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "To solve this problem, [3] and [4] proposed methods of collecting knowledge for answering questions from FAQ documents and technical manuals by using the document structure, such as, a dictionary-like structure and if-then format description. However, these kinds of documents requires the considerable cost of developing and maintenance. As a result, it is important to investigate a method of extracting evidential sentences for answering how type questions from natural language documents at low cost. To solve this problem, we proposed a method of developing a knowledge base by using mails posted to a mailing list (ML) [8] . We have the following advantages when we develop knowledge base by using mails posted to a mailing list.", | |
| "cite_spans": [ | |
| { | |
| "start": 23, | |
| "end": 26, | |
| "text": "[3]", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 31, | |
| "end": 34, | |
| "text": "[4]", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 625, | |
| "end": 628, | |
| "text": "[8]", | |
| "ref_id": "BIBREF6" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "it is easy to collect question and answer mails in a specific domain, and there is some expectation that information is updated by participants Furthermore, we developed a QA system and show that mails posted to a mailing list can be used as a knowledge base by which a QA system answers how type questions [8] . Next problem is wrong information. It is almost inevitable that natural language documents, especially web documents, contain wrong information. For example, (DA3-1) is opposed by (QR3-1-1).", | |
| "cite_spans": [ | |
| { | |
| "start": 307, | |
| "end": 310, | |
| "text": "[8]", | |
| "ref_id": "BIBREF6" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "(Q3) How I set up my wheel mouse for the netscape ? (DA3-1) You can find a setup guide in the Dec. issue of SD magazine.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "(QR3-1-1) I cannot use it although I modified /usr/lib/netscape/ja/Netscape according to the guide.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "Wrong information is a central problem of developing a knowledge base by using natural language documents. As a result, it is important to investigate a method of detecting and correcting wrong information in natural language documents. In this paper, we first report a method of detecting wrong information in question and answer mails posted to a mailing list. In our method, wrong information in the mails are detected by using mails which ML participants submitted for correcting wrong information in the previous mails. Then, the system gives one of the following confirmation labels to each set of question and their answer mails:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "positive label shows the information described in a set of a question and its answer mail is confirmed by the following mails, negative label shows the information is opposed by the following mails, and other label shows the information is not yet confirmed.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "Our knowledge base is composed of these labeled sets of a question and its answer mail. Finally, we describe a QA system: It finds question mails which are similar to user's question and shows the results to the user. The similarity between user's question and a question mail is calculated by matching of user's question and the significant sentence extracted from the question mail. A user can easily choose and access information for solving problems by using the significant sentences and confirmation labels.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "There are mailing lists to which question and answer mails are posted frequently. For example, in Vine Users ML, several kinds of question and answer mails are posted by participants who are interested in Vine Linux 1 . We intended to use these question and answer mails for developing knowledge base for a QA system because it is easy to collect question and answer mails in a specific domain, -it is easy to extract reference relations among mails, -there is some expectation that information is updated by participants, and there is some expectation that wrong information in the previous mails is pointed out and corrected by participants.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "However, there is a problem of extracting knowledge from mails posted to a mailing list. As mentioned, it is difficult to extract knowledge for answering how type questions from natural language documents only by using linguistic clues, such as, common content words and phrases. To solve this problem, [3] and [4] proposed methods of collecting knowledge from FAQ documents and technical manuals by using the document structure, such as, a dictionary-like structure and if-then format description. However, mails posted to a mailing list, such as Vine Users ML, do not have a firm structure because questions and their answers are described in various ways. Because of no firm structure, it is difficult to extract precise information from mails posted to a mailing list in the same way as [3] and [4] did. However, a mail posted to ML generally has a significant sentence. A significant sentence of a question mail has the following features: Before we discuss the significant sentence in answer mails, we classified answer mails into three types: (1) direct answer (DA) mail, (2) questioner's reply (QR) mail, and (3) the others. Direct answer mails are direct answers to the original question. Questioner's reply mails are questioner's answers to the direct answer mails. Suppose that (Q2) in Section 1 and its answers are question and answer mails posted to a mailing list, respectively. In this case, (DA2-1) and (DA2-2) are DA mails to (Q2). (QR2-2-1) is a QR mail to (DA2-2). (O2-1-1) and (O2-1-2) are the others.", | |
| "cite_spans": [ | |
| { | |
| "start": 303, | |
| "end": 306, | |
| "text": "[3]", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 311, | |
| "end": 314, | |
| "text": "[4]", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 791, | |
| "end": 794, | |
| "text": "[3]", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 799, | |
| "end": 802, | |
| "text": "[4]", | |
| "ref_id": "BIBREF2" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "In a DA mail, the answerer gives answers to the questioner, such as (DA2-1) and (DA2-2). Also, the answerer often asks the questioner back when the question is imperfect. As a result, significant sentences in DA mails can be classified into two types: answer type and question type sentence. They have the following features:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "it often includes the typical expressions, such as,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "\u2022 answer type sentence * dekiru / dekinai (can / cannot) * shita / shimashita / shiteimasu / shiteimasen (did / have done / doing / did not do) * shitekudasai / surebayoi (please do / had better)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "\u2022 question type sentence * masuka / masenka / desuka (did you / did not you / do you) -it is often quoted in the following mails.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "it often occurs after and near to the significant sentence of the question mail if it is quoted.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "In a QR mail, the questioner shows the results, conclusions, and gratitude to the answerers, such as (QR2-2-1), and sometimes points out wrong information in a DA mail and correct it, such as, (QR2-2-1) and (QR3-1-1). A significant sentence in a QR has the following features:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "it often includes the typical expressions.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "\u2022 dekita / dekimasen (could / could not)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "\u2022 arigatou (thank)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "it often occurs after and near to the significant sentence of the DA mail if it is quoted.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "Taking account of these features, we proposed a method of extracting significant sentences from question mails and their DA mails by using surface clues [8] . Then, we showed, by using the significant sentences extracted from question and their DA mails, the system can answer user's questions or, at least, give a good hint to the user. In this paper, we show that wrong information in a set of a question mail and its DA mail can be detected by using the QR mail. Then, we examined whether a user can easily choose and access information for solving problems with our QA system. In the next section, we will explain how to extract significant sentences from QR mails by using surface clues and confirm information in a set of a question mail and its DA mail.", | |
| "cite_spans": [ | |
| { | |
| "start": 153, | |
| "end": 156, | |
| "text": "[8]", | |
| "ref_id": "BIBREF6" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Mails Posted to a Mailing List", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "Information in a set of a question and its DA mail is confirmed by using the QR mail in the next way: step 1. extract a question mail, and its DA and QR mails by using reference relations and sender's email address. step 2. extract sentences from each mail by detecting periods and blank lines. step 3. check each sentence whether it is quoted in the following mails. step 4. extract the significant sentence from the question mail by using surface clues, such as, words in the subject, quotation in the DA mails, and clue expressions in the same way as [8] did. step 5. extract the significant sentence from the DA mail by using surface clues, such as, quotation in the QR mail, and clue expressions in the same way as [8] did. step 6. calculate the significant score of each sentence in the QR mail by applying the next two rules. The sentence which has the largest score is selected as the significant sentence in the QR mail. rule 6-1: a rule for typical expressions. Give n points to sentences which include n clue expressions in Figure 1 . rule 6-2: when two or more sentences have the largest score by applying rule 6-1, (1) give 1 point to the sentence which is located after and the nearest to the significant sentence in the DA mail if it is quoted, or (2) give 1 point to the sentence which is the nearest to the lead. step 7. give one of the following confirmation labels to the set of the question and DA mail. positive label is given to the set of the question and its DA mail when the significant sentence in the QR mail has type 1 clue expressions in Fig 1. negative label is given to the set of the question and its DA mail when the significant sentence in the QR mail has type 2 clue expressions in Fig 1. other label is given to the set of the question and its DA mail when the significant sentence in the QR mail has neither type 1 nor type 2 clue expressions in Fig 1. type For evaluating our method, we selected 100 examples of question mails in Vine Users ML. They have 121 DA mails, each of which has one QR mail.", | |
| "cite_spans": [ | |
| { | |
| "start": 554, | |
| "end": 557, | |
| "text": "[8]", | |
| "ref_id": "BIBREF6" | |
| }, | |
| { | |
| "start": 720, | |
| "end": 723, | |
| "text": "[8]", | |
| "ref_id": "BIBREF6" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 1035, | |
| "end": 1043, | |
| "text": "Figure 1", | |
| "ref_id": "FIGREF0" | |
| }, | |
| { | |
| "start": 1567, | |
| "end": 1573, | |
| "text": "Fig 1.", | |
| "ref_id": "FIGREF0" | |
| }, | |
| { | |
| "start": 1717, | |
| "end": 1723, | |
| "text": "Fig 1.", | |
| "ref_id": "FIGREF0" | |
| }, | |
| { | |
| "start": 1883, | |
| "end": 1889, | |
| "text": "Fig 1.", | |
| "ref_id": "FIGREF0" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "First, we examined whether the results of determining the confirmation labels were good or not. The results are shown in Table 1 . Table 2 shows the type and number of incorrect confirmation. The reasons of the failures were as follows:", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 121, | |
| "end": 128, | |
| "text": "Table 1", | |
| "ref_id": "TABREF2" | |
| }, | |
| { | |
| "start": 131, | |
| "end": 138, | |
| "text": "Table 2", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "there were many significant sentences which did not include the clue expressions. -there were many sentences which were not significant sentences but included the clue expressions. -some question mails were submitted not for asking questions, but for giving some news, notices, and reports to the participants. In these cases, there were no answer in the DA mail and no sentence in the QR mail for confirming the previous mails. -questioner's answer was described in several sentences and one of them was extracted, and misspelling.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Next, we examined whether these significant sentences and the confirmation labels were helpful in choosing and accessing information for solving problems. Our QA system put the significant sentences in reference order, such as, (Q4) vedit ha, sonzai shinai file wo hirakou to suru to core wo haki masuka.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "(Does vedit terminate when we open a new file?) (DA4-1) hai, core dump shimasu. (Yes, it terminates.) (DA4-2) shourai, GNOME ha install go sugu tsukaeru no desu ka?", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "(In near future, can I use GNOME just after the installation?)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Then, we examined whether a user can easily choose and access information for solving problems. In other words, we examined whether there was good connection between the significant sentences or not, and the confirmation label was proper or not.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "For example, (Q4) and (DA4-1) have the same topic, however, (DA4-2) has a different topic. In this case, (DA4-1) is a good answer to question (Q4). A user can access the document from which (DA4-1) was extracted and obtain more detailed information. As a result, the set of (Q4) and (DA4-1) was determined as correct. On the contrary, the set of (Q4) and (DA4-2) was a failure. In this experiment, 87 sets of a question and its DA mail were determined as correct and 34 sets were failures. The reasons of the failures were as follows:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "wrong significant sentences extracted from question mails, and wrong significant sentences extracted from DA mails.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Failures which were caused by wrong significant sentences extracted from question mails were not serious. This is because there is not much likelihood of matching user's question and wrong significant sentence extracted from question mails. On the other hand, failures which were caused by wrong significant sentences extracted from DA mails were serious. In these cases, significant sentences in the question mails were successfully extracted and there is likelihood of matching user's question and the significant sentence extracted from question mails. Therefore, the precision of the significant sentence extraction was emphasized in this task. Next, we examined whether proper confirmation labels were given to these 87 good sets of a question and its DA mail or not, and then, we found that proper confirmation labels were given to 64 sets in them. The result was shown in Table 3 .", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 879, | |
| "end": 886, | |
| "text": "Table 3", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "We discuss some example sets of significant sentences in detail. Question (Q5) in Figure 2 has two answers, (DA5-1) and (DA5-2). (DA5-1) is a suggestion to the questioner of (Q5) and (DA5-2) explains answerer's experience. The point to be noticed is (QR5-1-1). Because (QR5-1-1) contains type 1 expression in Figure 1 , it gives a positive label to the set of (Q5) and (DA5-1). It guarantees the information quality of (DA5-1) and let the user choose and access the answer mail from which (DA5-1) was extracted. Example (Q6) is an interesting example. (DA6-1) in Figure 2 which was extracted from a DA mail has wrong information. Then, the questioner of (Q6) confirmed whether the given information was helpful or not, and then, posted (QR6-1-1) in order to point out and correct the wrong information in (DA6-1) . In this experiment, we found 16 cases where the questioners posted reply mails in order to correct the wrong information, and the system found 10 cases in them and gave negative labels to the sets of the question and its DA mail. Figure 3 shows the overview of our system. A user can ask a question to the system in a natural language. Then, the system retrieves similar questions from mails posted to a mailing list, and shows the user the significant sentences which were extracted from the similar question and their answer mails. A user can easily choose and access information for solving problems by using the significant sentences and the confirmation labels. The system consists of the following modules:", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 82, | |
| "end": 90, | |
| "text": "Figure 2", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 309, | |
| "end": 317, | |
| "text": "Figure 1", | |
| "ref_id": "FIGREF0" | |
| }, | |
| { | |
| "start": 563, | |
| "end": 571, | |
| "text": "Figure 2", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 805, | |
| "end": 812, | |
| "text": "(DA6-1)", | |
| "ref_id": "FIGREF0" | |
| }, | |
| { | |
| "start": 1045, | |
| "end": 1053, | |
| "text": "Figure 3", | |
| "ref_id": "FIGREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Confirmation of Question and Answer Mails Posted to ML", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "question and answer mails (50846 mails), -significant sentences (26334 sentences: 8964, 13094, and 4276 sentences were extracted from question, DA, and QR mails, respectively), -confirmation labels (4276 labels were given to 3613 sets of a question and its DA mail), and synonym dictionary (519 words). Input analyzer transforms user's question into a dependency structure by using JUMAN [6] and KNP [5] .", | |
| "cite_spans": [ | |
| { | |
| "start": 388, | |
| "end": 391, | |
| "text": "[6]", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 400, | |
| "end": 403, | |
| "text": "[5]", | |
| "ref_id": "BIBREF3" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Similarity calculator calculates the similarity between user's question and a significant sentence in a question mail posted to a mailing list by using their common content words and dependency trees in the next way:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The weight of a common content word t which occurs in user's question Q and significant sentence S i in the mails", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "M i (i = 1 \u2022 \u2022 \u2022 N ) is: w W ORD (t, M i ) = tf (t, S i ) log N df (t)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "where tf (t, S i ) denotes the number of times content word t occurs in significant sentence S i , N denotes the number of significant sentences, and df (t) denotes the number of significant sentences in which content word t occurs. Next, the weight of a common modifier-head relation in user's question Q and significant sentence S i in question mail M i is:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "w LIN K (l, M i ) = w W ORD (modif ier(l), M i ) + w W ORD (head(l), M i )", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "where modif ier(l) and head(l) denote a modifier and a head of modifierhead relation l, respectively. Therefore, the similarity score between user's question Q and significant sentence S i of question mail M i , SCORE(Q, M i ), is set to the total weight of common content words and modifier-head relations which occur user's question Q and significant sentence S i of question mail M i , that is,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "SCORE(Q, M i ) = t\u2208Ti w W ORD (t, M i ) + l\u2208Li w LIN K (l, M i )", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "where the elements of set T i and set L i are common content words and modifier-head relations in user's question Q and significant sentence S i in question mail M i , respectively. When the number of common content words which occur in user's question Q and significant sentence S i in question mail M i is more than one, the similarity calculator calculates the similarity score and sends it to the user interface. User Interface. Users can access to the system via a WWW browser by using CGI based HTML forms. User interface put the answers in order of the similarity scores.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Knowledge Base. It consists of", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "For evaluating our method, we gave 32 questions in Figure 4 to the system. These questions were based on question mails posted to Linux Users ML. The result of our method was compared with the result of full text retrieval", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 51, | |
| "end": 59, | |
| "text": "Figure 4", | |
| "ref_id": "FIGREF3" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Evaluation", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "(1) I cannot get IP address again from DHCP server.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Evaluation", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "(2) I cannot make a sound on Linux.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Evaluation", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "(3) I have a problem when I start up X Window System. (4) Tell me how to restore HDD partition to its normal condition. Table 4 (a) shows the number of questions which were given the proper answer. Table 4 (b) shows the number of proper answers. Table 4 (c) shows the number and type of confirmation labels which were given to proper answers. In Test 1, our system answered question 2, 6, 7, 8, 13, 14, 15, 19, and 24. In contrast, the full text retrieval system answered question 2, 5, 7, 19, and 32. Both system answered question 2, 7 and 19, however, the answers were different. This is because several solutions of a problem are often sent to a mailing list and the systems found different but proper answers. In all the tests, the results of our method were better than those of full text retrieval. Our system answered more questions and found more proper answers than the full text retrieval system did. Furthermore, it is much easier to choose and access information for solving problems by using the answers of our QA system than by using the answers of the full text retrieval system.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 120, | |
| "end": 127, | |
| "text": "Table 4", | |
| "ref_id": "TABREF4" | |
| }, | |
| { | |
| "start": 198, | |
| "end": 205, | |
| "text": "Table 4", | |
| "ref_id": "TABREF4" | |
| }, | |
| { | |
| "start": 246, | |
| "end": 253, | |
| "text": "Table 4", | |
| "ref_id": "TABREF4" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Evaluation", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "Both systems could not answer question 4, \"Tell me how to restore HDD partition to its normal condition\". However, the systems found an answer in which the way of saving files on a broken HDD partition was mentioned. Interestingly, this answer may satisfy a questioner because, in such cases, our desire is to save files on the broken HDD partition. In this way, it often happens that there are gaps between what a questioner wants to know and the answer, in several aspects, such as concreteness, expression and assumption. To overcome the gaps, it is important to investigate a dialogue system which can communicate with the questioner.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Evaluation", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "Vine Linux is a linux distribution with a customized Japanese environment.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "NII-NACSIS Test Collection for IR Systems", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "TREC (Text REtrieval Conference) : http://trec.nist.gov/ 2. NTCIR (NII-NACSIS Test Collection for IR Systems) project: http://research.nii.ac.jp/ntcir/index-en.html", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Dialogue Helpsystem based on Flexible Matching of User Query with Natural Language Knowledge Base, 1st ACL SIGdial Workshop on Discourse and Dialogue", | |
| "authors": [ | |
| { | |
| "first": "Higasa", | |
| "middle": [], | |
| "last": "Kurohashi", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "141--149", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kurohashi and Higasa: Dialogue Helpsystem based on Flexible Matching of User Query with Natural Language Knowledge Base, 1st ACL SIGdial Workshop on Discourse and Dialogue, pp.141-149, (2000).", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Dialog Navigator\" A Question Answering System based on Large Text Knowledge Base, 19th COLING (COLING02)", | |
| "authors": [ | |
| { | |
| "first": "Kurohashi", | |
| "middle": [], | |
| "last": "Kiyota", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kido", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "", | |
| "volume": "8", | |
| "issue": "", | |
| "pages": "460--466", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kiyota, Kurohashi, and Kido: \"Dialog Navigator\" A Question Answering System based on Large Text Knowledge Base, 19th COLING (COLING02), pp.460-466, (2002.8).", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "A syntactic analysis method of long Japanese sentences based on the detection of conjunctive structures", | |
| "authors": [ | |
| { | |
| "first": "Nagao", | |
| "middle": [], | |
| "last": "Kurohashi", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1994, | |
| "venue": "Computational Linguistics", | |
| "volume": "20", | |
| "issue": "4", | |
| "pages": "507--534", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kurohashi and Nagao: A syntactic analysis method of long Japanese sentences based on the detection of conjunctive structures, Computational Linguistics, 20(4),pp.507-534, (1994).", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Helpdesk-oriented Question Answering Focusing on Actions", | |
| "authors": [ | |
| { | |
| "first": "Fujii", | |
| "middle": [], | |
| "last": "Mihara", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ishikawa", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1096--1099", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Mihara, fujii, and Ishikawa: Helpdesk-oriented Question Answering Focusing on Actions (in Japanese), 11th Convention of NLP, pp. 1096-1099, (2005).", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "A Question Answer System Using Mails Posted to a Mailing List", | |
| "authors": [ | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Watanabe", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yokomizo", | |
| "middle": [], | |
| "last": "Sono", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Okada", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "67--73", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Watanabe, Sono, Yokomizo, and Okada: A Question Answer System Using Mails Posted to a Mailing List, ACM DocEng 2004, pp.67-73, (2004).", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "num": null, | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "Clue expressions for extracting a significant sentence from a QR mail" | |
| }, | |
| "FIGREF1": { | |
| "num": null, | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "System overview QA processor. It consists of input analyzer and similarity calculator." | |
| }, | |
| "FIGREF2": { | |
| "num": null, | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "Where is the configuration file for giving SSI permission to Apache ? (6) I cannot login into proftpd. (7) I cannot input kanji characters. (8) Please tell me how to build a Linux router with two NIC cards. (9) CGI cannot be executed on Apache 1.39. (10) The timer gets out of order after the restart.(11) Please tell me how to show error messages in English. (12) NFS server does not go. (13) Please tell me how to use MO drive. (14) Do you know how to monitor traffic load on networks. (15) Please tell me how to specify kanji code on Emacs. (16) I cannot input \\ on X Window System. (17) Please tell me how to extract characters from PDF files. (18) It takes me a lot of time to login. (19) I cannot use lpr to print files. (20) Please tell me how to stop making a backup file on Emacs. (21) Please tell me how to acquire a screen shot on X window. (22) Can I boot linux without a rescue disk? (23) Pcmcia drivers are loaded, but, a network card is not recognized. (24) I cannot execute PPxP. (25) I am looking for FTP server in which I can use chmod command. (26) I do not know how to create a Makefile. (27) Please tell me how to refuse the specific user login. (28) When I tried to start Webmin on Vine Linux 2.5, the connection to localhost:10000 was denied. (29) I have installed a video capture card in my DIY machine, but, I cannot watch TV programs by using xawtv. (30) I want to convert a Latex document to a Microsoft Word document. (31) Can you recommend me an application for monitoring resources? (32) I cannot mount a CD-ROM drive." | |
| }, | |
| "FIGREF3": { | |
| "num": null, | |
| "uris": null, | |
| "type_str": "figure", | |
| "text": "32 questions which were given to the system for the evaluation" | |
| }, | |
| "TABREF0": { | |
| "content": "<table/>", | |
| "html": null, | |
| "num": null, | |
| "type_str": "table", | |
| "text": "1. it often includes nouns and unregistered words which are used in the mail subject. 2. it is often quoted in the answer mails. 3. it often includes the typical expressions, such as, (a) (ga / shikasi (but / however)) + \u2022 \u2022 \u2022 + mashita / masen / shouka / imasu (can / cannot / whether / current situation is) + . (ex) Bluefish de nihongo font ga hyouji deki masen. (I cannot see Japanese fonts on Bluefish.) (b) komatte / torabutte / goshido / ? (have trouble / is troubling / tell me / ?) (ex) saikin xstart ga dekinakute komatte imasu (In these days, I have trouble executing xstart.) 4. it often occurs near the beginning." | |
| }, | |
| "TABREF2": { | |
| "content": "<table><tr><td colspan=\"4\">type correct incorrect total</td><td/></tr><tr><td colspan=\"2\">positive 35</td><td>18</td><td>53</td><td/></tr><tr><td colspan=\"2\">negative 10</td><td>4</td><td>14</td><td/></tr><tr><td>other</td><td>48</td><td>6</td><td>54</td><td/></tr><tr><td colspan=\"5\">Table 2. Type and number of incorrect confirmation</td></tr><tr><td colspan=\"5\">incorrect type and number of correct answers</td></tr><tr><td colspan=\"4\">confirmation positive negative other</td><td>total</td></tr><tr><td>positive</td><td>-</td><td>4</td><td>1 4</td><td>18</td></tr><tr><td>negative</td><td>2</td><td>-</td><td>2</td><td>4</td></tr><tr><td>other</td><td>4</td><td>2</td><td>-</td><td>6</td></tr><tr><td colspan=\"5\">Table 3. Results of determining confirmation labels to the proper sets of a question</td></tr><tr><td>and its DA mail</td><td/><td/><td/><td/></tr><tr><td colspan=\"5\">labeling result positive negative other total</td></tr><tr><td>correct</td><td>29</td><td>8</td><td>27</td><td>64</td></tr><tr><td>failure</td><td>4</td><td>4</td><td>1 5</td><td>23</td></tr></table>", | |
| "html": null, | |
| "num": null, | |
| "type_str": "table", | |
| "text": "Results of determining confirmation labels" | |
| }, | |
| "TABREF3": { | |
| "content": "<table><tr><td>(I tried 'sndconfig' and became happy.)</td></tr><tr><td>(Q6) ES1868 no sound card wo tsukatte imasu ga, oto ga ookisugite komatte</td></tr><tr><td>imasu. (My trouble is that sound card ES1868 makes a too loud noise.)</td></tr><tr><td>(DA6-1) xmixer wo tsukatte kudasai. (Please use xmixer.)</td></tr><tr><td>(QR6-1-1) xmixer mo xplaycd mo tsukaemasen.</td></tr><tr><td>(I cannot use xmixer and xplaycd, too.)</td></tr><tr><td>Fig. 2. Examples of the significant sentence extraction</td></tr></table>", | |
| "html": null, | |
| "num": null, | |
| "type_str": "table", | |
| "text": "(Q5) sound no settei de komatte imasu.(I have much trouble in setting sound configuration.) (DA5-1) mazuha, sndconfig wo jikkou shitemitekudasai.(First, please try 'sndconfig'.) (QR5-1-1) kore de umaku ikimashita. (I did well.) (DA5-2) sndconfig de, shiawase ni narimashita." | |
| }, | |
| "TABREF4": { | |
| "content": "<table><tr><td/><td/><td colspan=\"4\">Test 1 Test 2 Test 3</td></tr><tr><td colspan=\"2\">our method</td><td/><td>9</td><td>15</td><td>17</td></tr><tr><td colspan=\"4\">full text retrieval 5</td><td>5</td><td>8</td></tr><tr><td/><td colspan=\"5\">(a) the number of questions which</td></tr><tr><td/><td colspan=\"5\">were given the proper answer</td></tr><tr><td/><td/><td colspan=\"4\">Test 1 Test 2 Test 3</td></tr><tr><td colspan=\"2\">our method</td><td/><td>9</td><td>25</td><td>42</td></tr><tr><td colspan=\"4\">full text retrieval 5</td><td>9</td><td>15</td></tr><tr><td/><td colspan=\"5\">(b) the number of proper answers</td></tr><tr><td/><td colspan=\"5\">positive negative other positive & negative</td></tr><tr><td>Test 1</td><td>2</td><td>2</td><td>5</td><td/><td>0</td></tr><tr><td>Test 2</td><td>9</td><td>4</td><td>12</td><td/><td>0</td></tr><tr><td>Test 3</td><td>10</td><td>5</td><td>25</td><td/><td>2</td></tr><tr><td/><td colspan=\"5\">(c) the number and type of labels</td></tr><tr><td/><td colspan=\"4\">given to proper answers</td></tr><tr><td colspan=\"2\">Test 1. by examined first answer</td><td/><td/><td/></tr><tr><td colspan=\"4\">Test 2. by examined first three answers</td><td/></tr><tr><td colspan=\"3\">Test 3. by examined first five answers</td><td/><td/></tr></table>", | |
| "html": null, | |
| "num": null, | |
| "type_str": "table", | |
| "text": "Results of finding a similar question by matching of user's question and a significant sentence" | |
| } | |
| } | |
| } | |
| } |