Study of visualization method of committee minutes using text mining. A case study of waste management field

Размещено на http://www.allbest.ru/

Study of visualization method of committee minutes using text mining. a case study of waste management field

Koyama Fumitaka, Ishii K.

Hokkaido University, Sapporo, Japan

Abstract

In Japan, committees composed of many stakeholders are held for planning in waste management field. However, transparency of decision-making processes and disclosure of information are insufficient to form a consensus. There are few committees that have published minutes in the waste management field, and few studies have analyzed the minutes at the present time. Therefore, we attempted to visualize contents of minutes that are published as full-text version by text mining. In this study, we analyzed a case of “Aomori * Iwate Prefecture Illegal Dumping”, to propose a text mining method. Particularly, we selected unnecessary words and extracted important words to analyze whether each person in committee had specific and critical opinion on the discussion, by frequency of words that each person used for the opinion/synchronized expression. Through these analyses, participation degree into discussion, technical interest, and critical opinion on the discussion are clarified for each person in committee. This study will help visualization of contents of minutes to promote the discussion among persons in committee.

Key words: waste management, minutes of committee, text mining, information disclosure, attitude to discussion.

Абстракт

linguistic analysis text

Кояма Фумитака, Иши К.

УХ, г. Саппоро, Япония

ИССЛЕДОВАНИЕ МЕТОДА ВИЗУАЛИЗАЦИИ ПРОТОКОЛОВ ПУТЁМ АНАЛИЗА ТЕКСТА. ТЕМАТИЧЕСКОЕ ИССЛЕДОВАНИЕ В СФЕРЕ КОНТРОЛЯ НАД ОТХОДАМИ

В Японии существуют состоящие из множества заинтересованных сторон комитеты, занимающиеся разработками в сфере управления отходами. Однако открытость процесса принятия решений и разглашение информации о них недостаточны для общего согласия. В настоящее время всего несколько комитетов опубликовали протоколы по регулированию отходов, и небольшое количество исследований проанализировало эти протоколы. По этой причине мы попытались визуализировать содержание документов, опубликованных в полной версии, путём анализа текста.

В данной работе исследуется пример “Аомори. Незаконная свалка в префектуре Ивате”. Мы выделили ненужные фразы и извлекли важные, чтобы выяснить, имел ли каждый член комитета индивидуальное критическое мнение в отношении данной проблемы. Для этого потребовалось установить частоту используемых ими слов.

Таким образом, относительно каждого члена комитета были выяснены степень вовлеченности в обсуждение, технический интерес, наличие критического видения. Настоящая статья поможет визуализировать содержание протоколов, чтобы стимулировать продуктивные дискуссии внутри комитета.

Ключевые слова: контроль над отходами, протокол комитета, анализ текста, разглашение информации, отношение к обсуждению.

Introduction

There is a method called text mining of minutes as an objective analysis method of discussion. As past studies of text mining for minutes of committees in Japan, Iwami et al. (2014) analyzed the minutes of committee on a river project to visualize a relationship among the committee members [1]. In addition, Masuda (2012) analyzed minutes of the local administrations to clarify activities of assemblies [2]. However, there are few studies on text mining of minutes in waste management field. According to CiNii article search [3], just six articles are shown using "Waste + text mining" as search words. There are a few studies of questionnaire, but no studies of analysis minutes related to waste management field. Figure.1 shows current situation in publishment of committee minutes in waste management field in Japan (Fig.1).

Although committees have little disclosure in waste management field, as shown in Figure 1, this study attempted to analyze the committee whose minutes are published in full text. The minutes can be objectively visualized as in previous researches in other fields. The committees in waste management field handled various issues, such as siting for final disposal sites and incineration facilities, and restoration of illegal dumping sites. In addition, the committees composed of many stakeholders, such as not only administrators and experts but also residents, with diverse values and backgrounds. Decision making processes toward agreement for all concerned parties are vital. For that, it is indispensable to make the process transparent and to disclose all information.

Figure!. Current situation in publishment of committee minutes in waste management field in Japan

This study proposes a method to use the minutes of the committee effectively to promote transparency of decision-making processes and disclosure of information. Therefore, the objective of this study is to develop a method to objectively visualize contents of minutes by applying a text mining method to waste management field.

1. Methods

Target Proceedings: The minutes of "Technical working group on Aomori * Iwate Prefecture Illegal Dumping Site" whose minutes are published in waste management field were objective in this study. The total number of meeting was five. The minutes composed of (A) document explanation part by the administrators and (B) discussion part by the committee members. In this study, we extracted only the discussion part. A software called “ttm [4]” was used for text mining analysis. The number of character in the target words was two or more, and person names were excluded as the target words.

Selection of unnecessary words: We extracted the target word from the text data only for the discussion part by the committee members. The coefficient of variation (CV) for the extracted words was calculated using the frequency of word appearance in each part, as shown in equation (1). The words with the CV of lower than10% were designated as unnecessary words. The small coefficient of variation means that the word was not directly related to the content of discussion.

CV = oDF/DF(1)

Where, оis standard deviation, DF represents document frequency (Number of sentences including the word), and (DF) is the average of document frequency.

Extraction of important words: The set of unnecessary words were excluded from text data on only the discussion parts of the five minutes. We extracted important word candidates up to 1000 words in the order of DF. According to Document Frequency Inverse Unit Frequency (DFIUF) for the important word candidate, as shown in equation (2), importance of words in each discussion part was evaluated. Words with larger DFIUF mean more important than other words. The DFIUF evaluates words specifically appeared in a part of discussion because larger DFIUF means the words appeared more frequently in the part than in the other parts.

DFIUF = DFxlog(U/UF) (2)

Where, UF is unit frequency (Number of discussion part including the word), U: Units (total discussion part number in a time of committee)

Attitude of committee members for a basic plan of the restoration: We analyzed the following items: 1) The number of remarks spoken by the committee members, 2) the number of the important word appearance, and 3) the number of times using the assertion (criticism) / synchronization expressions.

In the Aomori-Iwate illegal dumping site, the important issues were removal of all waste and countermeasures for prevention of contaminant spreading through groundwater. So, we focused on the discussion part on this issue. There were two policies: Construction of vertical walls to prevent contaminants from spreading out of the site and removal of waste after the construction, which was proposed by Aomori Prefecture (Policy 1) and Removal of waste without prevention of contaminant spreading, which was proposed by Iwate Prefecture (Policy 2). We investigated whether a committee member was more critical for the above policy.

2. Result and discussion

Number of remarks spoken by the committee members: The number of remarks spoken by each committee member represents the degree of participation into the discussion on the two policies described above. We standardized the number of remarks in each discussion part. Figure 2 shows tabulated results on the number of remarks for each committee member. Regarding the degree of participation, committees B, C, D and E had remarks against Policy 1 more frequently than the other committee. Committee I had remarks against Policy 2 frequently (Fig. 2).

Figure 2. The number of remarks on the two policies

Number of the important word appearance:

The number of important word appearance represents the degree of technical interest. The number of the important words for each committee member was analyzed and was shown in tabulated graph (Fig. 3). Committees B, E and G seemed to have technical interest in Policy 1 and committee I was interested in Policy 2.

Figure 3. The number of important words mentioned on the “removal plan basic policy"

Number of times using the assertion (criticism) / synchronization expressions: Committee members, who did not use synonymous expressions ("Watashimo" means “me, too” in Japanese) but assertion expressions ("Watashi ha" means “I” in Japanese) was considered to have critical opinions to the policy. As shown in Table 1, committee B had critical opinions towards Policy 1. On the other hands, committee's F and H had critical opinions towards Policy 2. (Table 1)

Attitude of committee members for a basic plan of the restoration: Attitude towards policies was summarized from the above three analyses on the degree of participation in the discussion, the degree of technical interest and critical opinions to the policy. (Table 2)

Tablel. The number of assertions / synchronization expressions spoken by the committee members

Table 2. Characterization by attitude to policy

Chairperson A did not show any attitude towards the two policies. On the other hand, committee B was critical to Policy 1, suggesting that committee B agreed with the policy 2. Since Policy 1 referred removal of waste but not all waste, committee B might be critical on removal of not all waste in Policy 1. In addition, committees F and H has critical opinions toward Policy 2, although they had small number of remarks.

Conclusion

We proposed a method to visualize the minutes and applied the method to clarify the attitude of committee members to the policies from three points of view: 1) the degree of participation in the discussion, 2) technical interest, and 3) critical opinion.

The attitude of each committee member towards the policy will promote interpretation of the minutes. In this study, we analyzed the minutes on which the policies are different between the both prefectures. In addition, each prefecture prepared the material for explanation of each countermeasure, separately. For similar kinds of minutes, we can apply the same method as this study, and visualize committee attitude. Now, a method of directly analyzing the content of the discussion is only the analysis using the member of opinions / synchronization expressions. Further investigation is required.

References

1. Iwami, A., Ohno, T., Kimura, M., & Ide, S. Developing a Text Mining Technique to Identify Concerted or Opposed Relations among Committee Members in Public Work Planning Processes with Using the Processes' Minutes; Japan Society of Civil Engineers, 2014,70(6), II_249 p-II_256 p

2. Masuda, T. Text Mining Analysis on The Minutes of Local Assemblies - A Case Study on the Takasaki City Assembly -; Studies of regional policy, 2012, 15(1), 17p-31p

3. CiNii Articles. Retrieved January 31, 2018 from https://ci.nii.ac.jp/

4. TTM: TinyTextMiner P version. Retrieved January 31, 2018 from http://mtmr.jp/ttm/

Размещено на Allbest.ru