With the currently growing interest in the Semantic Web and Social Networking, personal metadata is coming to play an important role in the Web. This paper proposes a novel keyword extraction method to extract personal metadata from the Web. The proposed method is based on co-occurrence information of words. Our method extracts relevant keywords depending on the context of a person. Our experimental results show that extracted keywords are useful for personal metadata creation. We also discuss the annotation of personal metadata and application to the Semantic Web.
Considering personal metadata, we notice that a lot of information is contained in the Web pages. For example, imagine a researcher: that researcher’s information can be in an affiliation page, a conference page, an online paper, or even in a Weblog. In fact, we can expect that these pages contain a lot of personal metadata even including information that we would not expect to find. Therein, questions are: – What kind of personal metadata are in the Web? – What kind of Web page contains personal metadata? – How are extracted metadata applied to semantic annotation?
Considering these points, one of our research goals is to extract personal metadata from the Web and apply them to semantic annotation. As a preliminary report to achieve this goal, we propose a novel keyword extraction method to extract personal information from the Web.
The remainder of this paper is organized as follows: section 2 describes the proposed keyword extraction method using an actual example. In section 3, we show the extracted keywords and analyze them. In section 4, we discuss the annotation of personal metadata. Section 5 contains related works. Finally, we address future works and conclude this paper in section 7. 2 2.1
As an experimental attempt, we extracted the keywords of Program Committee members of SemAnnot 2004 Workshop (There are 28 members including chair persons). First, we need to acquire Web pages that contain information of respective committee members and their mutual relationships. A simple way of acquiring those Web pages is to use a search engine. It is reasonable to use a search engine because it can search many Web pages in less than a few seconds. It also tracks the temporal variance of the Web. In this experiment, we used Google5, which currently addresses data from 4 billion Web pages.
We first put each person’s full name to a search engine (name is quoted with double quotation such as “Siegfried Handschuh”) and retrieve documents related to each person. From the search result, we used the top 10 documents per person as the initial documents that might contain personal keywords.
The search result documents include not only html files but also other file types such as .pdf, .doc, .xls, .ppt. In this experiment, we used only html files. Furthermore, we did not use metadata indicators in an html file such as META tags and RDF. In the future, we are planning to use other file types along with html files that already have been attached metadata.
The html files, at to a maximum of 10 files per person, are acquired from the initial
documents of each person. They are pre-processed with html-tag deletion and
part-ofspeech tagging (POS). Then, the term set for keyword extraction is extracted from
preprocessed html files using the term extraction tool, Termex [
from POS data based on statistical information of conjunctions between parts of speech. Termex6 can also extract nominal phrases that include more than two nouns such as “Annotation tool”. After the whole procedure of extracting the term set, we extracted about 1000 terms per person on the average. The relevant keyword of each person is chosen from these terms. Figure 1 shows steps of the proposed keyword extraction. Because the term set includes both relevant and irrelevant terms for personal information, we need to evaluate the relevance of term as a personal keyword. This subsection explains the scoring method that gives relevance as a personal keyword to the term. Term relevance based on Co-Occurrence The simple approach to measure term relevance as a personal keyword is to use co-occurrence. In this paper, we define cooccurrence of two terms as term appearance in the same Web page. If two terms cooccur in many pages, we can say that those two have a strong relation and one term is relevant for another term. This co-occurrence information is acquired by the number of retrieved documents of a search engine result. For example, assume we are to measure the relevance of name N (e.g. “Siegfried Handschuh”) and term w (e.g. “Annotation”). Here, w is the term in the term set W extracted from the initial documents of the person named “N ” . We first put a query, “N and w”, to a search engine and obtain the number of retrieved documents that is denoted by jN and wj. We continuously apply a query, “N ” and “w”, and obtain the number of retrieved documents for each, jN j and jwj. Then, the relevance between the name N and the term w, denoted by r(N; w), is
approximated by the following Jaccard coefficient.
r(N; w) =
jN and wj jN j + jwj ¡ jN and wj This Jaccard coefficient captures the degree of co-occurrence of two terms by their mutual degree of overlap.
Keyword of person As described in a previous subsection, the term set of a person is extracted from various Web pages. Although the Web pages contain a person’s name in the text, each page may contain personal information in different contexts. For example, imagine that one person, named “Tom”, is both a researcher and a artist, we can expect that his name may appear not only in academic-related pages, but also in other pages related to his art activities. Even among his academic-related pages, there might be different pages depending on his acquaintances, affiliations, and projects. In this way, different Web pages reflect different contexts of a person. Here, we introduce the notion of a context to extract the keyword that captures the context of a person.
To extract the keyword in relation to a certain context, we must estimate the relevance between the term and the context. If we replace the name N with the context C in the relevance, r(N; w), we can obtain the relevance between context C and term w, r(C; w), in the same manner. Then, the relevance of person N and term w in the context C, denoted by score(N; C; w), is calculated as the following.
score(N; C; w) =
r(N; w) M AX(r(N; w)) + ® Therein, ® denotes the relevance between the person and the context. For example, we can use r(N; C) as ®. M AX(r(X; Y )) is the maximum value of the Jaccard coefficient in the term set W . We define the threshold for r(N; C) to exclude terms that are not relevant for a person, but that have strong relation to the context. threshold is decided based on heuristic method. The term w with the higher score(N; C; w) is considered to be a more relevant keyword for person N in context C.
Regarding the “Tom” example, if we set “Art” as the context, we can get keywords related to his art activities. Alternatively, if we include his research project name as the context, keywords related to his project would be acquired.
two persons in terms of their contexts, one person can be regarded as a part of the context of another person. Hence, we can apply the previous formula to keyword extraction of the relations among persons as follows: score(N 1; N 2; W ) =
r(N 1; w) M AX(r(N 1; w)) + ¯
r(N 2; w)
M AX(r(N 2; w)) ( r(N 1; w) r(N 2; w)
; M AX(r(N 1; w) M AX(r(N 2; w)
> threshold) Therein, N 1 and N 2 denote each person’s names in the relation. ¯ is the parameter of relevance between persons, such as r(N 1; N 2). This formula shows the relevance of person N 1’s term w in relation to person N 2.
As there are many contexts of a person, the relations among persons also have a variety of contexts. For example, the relation of two persons in the academic field might be coauthors, have the same affiliation, the same project; they may even be friends. The relevance of person N 1’s term w in relation to person N 2 in the context C, score(N 1; N 2; C; w), is given as follows: score(N 1; N 2; C; w) = score(N 1; N 2; w) + ° Therein, ° is the parameter of relevance between the persons and the context, such as r(N 1andN 2; C).
As an example of extracted keywords, Table. 1 shows higher-ranked extracted keywords of “Siegfried Handschuh”. Each column in the table shows higher-ranked keywords based on Term Frequency Inverse Document Frequency (TFIDF), co-occurrence without the context, and co-occurrence with the context, respectively, from the left column.
In TFIDF-based keywords, we can find keywords that are related to the person such as “annotation” and “semantic”. Nevertheless, there are many irrelevant words including general words. Because TFIDF is based on the frequency of word appearances in a text, it is difficult for a word to become higher-ranked in terms of relevance with another word. On the other hand, in co-occurrence-based keywords, general words are excluded and relevant words of each person appear in the rank list.
As explained in the previous section, the context can be considered in the keyword extraction. In this experiment, we used “Semantic Web” as the context. With this context, keywords are chosen in relation to one’s activity about the Semantic Web. In the column of “Co-Occurrence with the context”, we can find that context-related keywords come to appear in the rank list. The order of higher-ranked keywords also changes in relation to the context.
The column at the right side shows a property label for each keyword in “CoOccurrence with the context”. Considering a correspondence to existing personal metadata vocabularies such as FOAF, we have defined six property labels: Name (N), Technical term (T), Event (E), Organization (O), Project (P), URL. In order to analyze what kind of property is included in keywords, we annotated a property label to higher-ranked keywords of each person. Thereby, we acquired 1646 labeled keywords in total (about 60 keywords per person on average).
Table. 2 shows the distribution of property labels. Nearly half of higher-ranked keywords are occupied with names. Notwithstanding, it is noteworthy that other properties such as organizations and projects also appear to a certain degree. In particular, as shown on the right side column, the properties for each person are distributed in a balanced manner. This distribution indicates that if we extract about 60 higher-ranked keywords of one person, on average we can obtain about 30 names of his acquaintance, 2 or 3 related organizations, and 1 or 2 projects. These numbers nearly match our research activity and show the possibility of using keywords for personal metadata. In this analysis, we took many keywords together as “technical terms”. If we classify each keyword more precisely, we could discover other personal metadata in keywords. To further explore the possibility of personal metadata extraction from the Web, we analyzed which Web pages include a higher-ranked keyword. First, we classified all 280 Web pages (10 per person) that were used to extract the initial term set. Thereby, we prepared the 11 categories shown in Table. 3. “Personal page” includes personal Web pages of the affiliation or one’s own domain. “Other page” includes uncategorized pages and non-html pages such as .pdf and .ppt files. “Event page” includes conference, workshop, and meeting pages. ML log is the email exchanged in a mailing list. DBLP 7 is the online bibliography of Computer Science papers. As seen in the table, “Personal page” is the most dominant type of Web page. Because a person’s name was used as a query, it is natural that we obtain a personal page in a search result.
Property
Number
Per person
Although name entities can be acquired most from “Personal page”, DBLP is also a good information resource to extract a name entity. DBLP contains coauthor information of a paper. Therefore, the extracted name is related to one’s acquaintance in a research activity. “Event page”, such as conference, workshop, is a information resource of various personal information. However, because the Event page is not specified to a certain person, “Personal page” gives more accurate information about each person.
Overall, “Personal page” is a good information source for personal metadata such as names, organizations, and projects. Event page and DBLP provide metadata that are related to personal research activities such as coauthors, projects, and events including conferences and workshops. 4
Our keyword extraction method can be applied to semantic annotation in following ways.
– Annotation for Web page : As our analysis showed, our personal keyword
extraction method offers strong potential for personal metadata extraction from the Web.
Extracted personal metadata can be applied to partially annotate the Web pages
using metadata description framework such as the RDF[
This paper presents discussion of the importance of a person’s context in keyword
extraction. The context often defines the properties. Currently, there is no FOAF
vocabulary to define a context. In addition to FOAF, there are many vocabularies and
framework for personal metadata such as Topicmaps [
Aiming at extracting and annotating personal metadta, our method is regarded as one
of Information Extraction(IE) methods supporting a semantic annotation. Up to now,
many IE methods rely on predefined templates and linguistic rules or machine learning
techniques to identify certain entities in text documents[
Some previous IE researches have addressed the extraction and annotation of
personal metadata. In [
In [
Although the aim is not extracting personal metadata, in [
To apply our keyword extraction methods to personal metadata annotation, we must consider and solve following points in the future.
– Evaluation of personal metadata : One problem is that we are not sure that the
extracted metadata are true. Although two terms co-occur in many Web pages, they
might not have any relation. Therefore, someone should evaluate the propriety of
a keyword as actual metadata. One approach to solve this problem would be an
interactive annotation system[
The Web holds much personal information that can be used as personal metadata. This paper proposes a novel keyword extraction method to extract personal information from the Web. Our result showed the important possibility of using extracted keywords as personal metadata. Importantly, our method can capture the personal information in different contexts. This allows us to obtain various personal metadata.
Because the Web is such a large information resource, its information runs the gamut from useful to trivial. It presents the limitation that it must be publicly available on the Web. For further improvement of the proposed method, we must analyze “what” information of “who” in the Web, and its reliability.