A corpus-based approach to automatic compound extraction

Abstract

An automatic compound retrieval method is proposed to extract compounds within a text message. It uses n-gram mutual information, relative frequency count and parts of speech as the features for compound extraction. The problem is modeled as a two-class classification problem based on the distributional characteristics of n-gram tokens in the compound and the non-compound clusters. The recall and precision using the proposed approach are 96.2% and 48.2% for bigram compounds and 96.6% and 39.6% for trigram compounds for a testing corpus of 49,314 words. A significant cutdown in processing time has been observed.

Measuring Termhood in Automatic Terminology Extraction

“…It is obvious that the percentage of shorter Model is reasonable because most new terms are Figure 5 shows the result of the SVM using the compounded terms from existing terms as third training set. This set uses a smaller domain mentioned in (Keh-Yih Su et al, 1994). It is lexicon, but a large training corpus data.…”

Tree pattern expression for extracting information from syntactically parsed text corpora

“…Especially, our system resolves certain technical variations of a PN (e.g. "TGF-α, TGF-alpha, TGF alpha, TGF-a, TGFa", "IGF-2, IGF2, IGF-II, IGF II") using a simple statistical method (Su et al 1994), and it can convert each PN within bio-texts to its representative acronym using an acronym-synonyms list, each of which was connected to its related entry in the PN dictionary. Our dictionary-based tagger was embedded in the Stanford NLP parser (Klein and Manning 2002) so that each sentence of the original texts in a corpus was parsed and represented as its list form (which, henceforth, we call the "bio-parse tree") with two additional taggers (one, ||, for the PN and the other, || ||, for the IV) as in the following examples.…”

Biological relation extraction and query answering from MEDLINE abstracts using ontology-based text mining

“…The NB method performs better on gene names (84%) while the DT method yields better results on protein names (85%) and other categories. Su et al [17] have proposed a corpus-based approach for automatic compound extraction, which considers only bigrammes and trigrams. However, there are instances of medical and biological entities containing up to seven words, and bigram or trigram-based approaches fail in such situations.…”