Turney, P. (2000), `Learning algorithms for keyphrase extraction', Information Retrieval 2(4), 303--336.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....that keyword indexing apart from being useful on its own may play a complementary role to full text indexing. In addition, it is an interesting task for machine learning experiments due to the complexity of the activity. We extend previous work on automatic keyword assignment (see e.g. [1]) to include knowledge from a thesaurus. In this article, we present experiments where we for each document in a collection automatically extract a list of potential keywords. This list, we envision, can be given to a human indexer, who in turn can choose the most suitable terms from the list. ....

.... take into account abbreviations, as some words are only represented by the short form in the text, e.g. IT, not by the full form informationsteknik (information technology) For applications to real life indexing tasks, some form of utility likelihood measure should be used in result (as in e.g. [1]) As stated earlier, we have so far only looked at single word terms. This means that a certain amount of both potential index terms as well as terms selected by the human indexers have been ignored. However, as Swedish is rich in compounding, this is much less of a problem than had it been for ....

Turney, P.D. (2000). Learning Algorithms for Keyphrase Extraction. Information Retrieval, 2(4):303--336. Kluwer Academic Publishers.

....first tags the input by assigning syntactic classes to each word. We use the Brill tagger [1,2] Then we experimented with two heuristics for noun Figure 6: Browsing for information on poisson Figure 5: Browsing for information on dairy phrase identification. The first was suggested by Turney [18] as matching almost all of the keyphrases in the corpuses he used. It specifies zero or more nouns or adjectives, followed by one final noun or gerund: noun adjective) noun verb gerund) where a noun is either a singular or plural noun or proper noun. means repetition, appearing zero ....

Turney, P.D. (in press) Learning algorithms for keyphrase extraction.Information Retrieval.

....to produce the best extracted summaries. To increase flexibility in the configurability of the system, we would ideally have a number of different modules that could be plugged in at the appropriate points in the greater system. For keyphrase extraction, we are using Peter Turney s Extractor [10, 11]. As an alternative we decided to build a simple keyphrase extractor in house as well. The goal was to keep the extractor simple and to apply any linguistic insight we might have to the process. This paper presents our simple keyphrase extractor (herein referred to as B C for lack of a better ....

....the costs involved in experiments involving human judges limit the scope of experimental evaluation. Our experiments, therefore, are restricted to a comparison of B C and Extractor on a small number of documents. Comparisons of Extractor to other keyphrase extraction systems can be found in [11]. As usual, many other experiments can be imagined and should be carried out (see section 7) The experiments suggest that B C and Extractor perform differently, but about equally well. Our judges preferred individual keyphrases from Extractor more often but complete sets from B C more often. A ....

[Article contains additional citation context not shown here]

Turney, Peter D. (2000). "Learning Algorithms for Keyphrase Extraction." Information Retrieval. To appear.

....performed tests on the TREC 5 collection . Note that all these works focus on ad hoc retrieval and rely on the adaptation of classical IR document ranking techniques for taking into account text passages instead of whole documents. Text segmentation has also been considered from an IE perspective. [21] uses decision trees for the extraction of keywords or key phrases (two or more words) from text, 23rd BCS European Annual Colloquium on Information Retrieval, 2001 2 HMM based Passage Models for Document Classification and Ranking framing the extraction problem as a classification problem on ....

Turney P. D., Learning Algorithms for Keyphrase Extraction, Information Retrieval, 2(4):303-336, 2000

....a structure on the collection by extracting important phrases from each document, but in Keyphind the documents are much longer and furthermore a manually tagged training corpus is needed to build the classifier. Extractor is a similar system for extracting keyphrases using supervised learning [34, 35]. Clustering is also being used for concept based relevance feedback for Web information retrieval [6] Following a user query the retrieved documents are organised into conceptual groups. Unlike in our approach this structure is not extracted for the indexed domain but for the search results. ....

P. D. Turney. Learning Algorithms for Keyphrase Extraction. Information Retrieval, 2(4):303--336, 2000.

....frequency) which corresponds to detecting words that not only have high frequency in the current document but also are relatively rare in a large collection of documents. In [Lin 95] the central concepts of the text were obtained through the generalization relationships found in WordNet. Turney 00] proposed the system GenEx, formed by two components: the genetic algorithm Genitor, which maximizes the performance in training data, and the algorithm Extractor, which obtains a list of key phrases from a document. In our system we propose a simpler idea: we restrict the analysis of the text ....

.... probability are selected for the summary; In the case of C4.5 which in its default form outputs only the predicted class, and not the class probabilities we have used the p option of this tool, which generates soft threshold decision trees providing an estimate of the class probabilities [Turney 00] Our system selects the n sentences with largest value of this probability to be included in the summary. Another problem we had to deal with in our experiments was the problem of unbalanced classes. In our case, only 10 or 20 of the examples belonged to the positive class (corresponding to a ....

Turney, P.D. Learning algorithms for keyphrase extraction. Information Retrieval, 2 (4). 2000.

No context found.

Turney, P.D. Learning algorithms for keyphrase extraction. Information Retrieval, 2, 303-336, 2000.

....of human labour. A major limitation of such hand generated lexicons is the relatively poor coverage of technical and scientific terms. For example, I am interested in applying synonym recognition algorithms to the Lecture Notes in Computer Science 5 automatic extraction of keywords from documents [15]. In a large collection of scien tific and technical journals, I found that only about 70 of the authors keywords were in WordNet. On the other hand, 100 were indexed by AltaVista. This is a strong motivation for automating aspects of the construction of lexical databases. Another motivation ....

....terms, because the top retrieved documents constitute a relatively small, noisy sample. Thus there could be some benefit to validating the suggested expansions using PMI IR, which would draw on larger sample sizes. I am particularly interested in applying PMI IR to automatic keyword extraction [15]. One of the most helpful clues that a word (or phrase) is a keyword in a given document is the frequency of the word. However, authors often use synonyms, in order to avoid boring the reader with repetition. This is courteous for human readers, but it complicates automatic keyword extraction. I ....

Turney, P.D.: Learning Algorithms for Keyphrase Extraction. Information Retrieval, 2 (2000) 303-336.

No context found.

Turney, P. (2000), `Learning algorithms for keyphrase extraction', Information Retrieval 2(4), 303--336.

No context found.

Turney, P. (2000). "Learning Algorithms for Keyphrase Extraction". In Information Retrieval 2(4), pp. 303-336.

No context found.

Turney, P.: Learning Algorithms for Keyphrase Extraction. In Information Retrieval 2(4), 303-336.

No context found.

Turney, P. Learning Algorithms for Keyphrase Extraction. In Information Retrieval 2(4), 303-336.

No context found.

Turney, P.D. Learning Algorithms for Keyphrase Extraction. Information Retrieval 2, 4 (2000); 303-336.

No context found.

Turney P. D., Learning Algorithms for Keyphrase Extraction, Information Retrieval, 2(4):303-336, 2000

No context found.

Turney, P.D. (in press) "Learning algorithms for keyphrase extraction." Information Retrieval.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC