Alexander Gelbukh

The Need for Full Co-Occurrence

We have previously shown that simultaneously considering three arguments yields better precision than does considering only two, though with a certain loss of recall. Kawahara and Kurohashi [107] differentiate the main verb using the closest argument in order to disambiguate verbs for learning preferences.

First Approach: Sentence Analysis Using Rewriting Rules

Some of the earliest useful user interaction systems involving sentence analysis used rewriting rules. A famous example of one such system is SHRDLU, which was created in the 1960s by Terry Winograd at the Massachusetts Institute of Technology and was able to solve many of the problems that arise when conversing with a computer.

Evaluation of the Dependency Parser

Many corpora are annotated using constituent formalism. However, our goal is to evaluate parsers within the dependency formalism, which means we need a gold standard in the dependency formalism.

Second Approach: Constituent Grammars

In this chapter, we tackle sentence analysis using the constituent approach. This approach has two problems. The first is the difficulty of extracting information about characters and actions from factual reports such as news articles, web pages, and circumscribed stories. To construct the structure of a sentence, there should be interaction with previously acquired knowledge. In turn, such knowledge should be expressed in a structured way so that simple inferences may be used when necessary. We will deal with this problem in detail in Sect. 3.1.

Third Approach: Dependency Trees

After exploring several approaches and representational structures in the previous two chapters, we found that the formalism that best suits our needs is the dependency tree representation. Thus, in this chapter, we present a parser that is based on a dependency tree. This parser’s algorithm uses heuristic rules to infer dependency relationships between words, and it uses word co-occurrence statistics (which are learned in an unsupervised manner) to resolve ambiguities such as PP attachments. If a complete parse cannot be produced, a partial structure is built with some (if not all) dependency relations identified.

Multiple Argument Handling

A sentence can be regarded as a verb with multiple arguments. The plausibility of each argument depends not only on the verb but also on other arguments. Measuring the plausibility of verb arguments is necessary in several tasks, such as semantic role labeling, where grouping verb arguments and measuring the plausibility increases performance [70, 135]. Metaphor recognition also requires knowledge of verb argument plausibility in order to recognize uncommon usages, which would suggest either the presence of a metaphor or a coherence mistake (e.g., drink the moon in a glass). Malapropism detection can use the measure of the plausibility of an argument to determine word misuse [185]—such as in hysteric center instead of historic center, density has brought me to you instead of destiny has brought me to you, a tattoo subject instead of a taboo subject, and don’t be ironing instead of don’t be ironic.

Adam Kilgarriff’s Legacy to Computational Linguistics and Beyond

The 2016 CICLing conference was dedicated to the memory of Adam Kilgarriff who died the year before. Adam leaves behind a tremendous scientific legacy and those working in computational linguistics, other fields of linguistics and lexicography are indebted to him. This paper is a summary review of some of Adam’s main scientific contributions. It is not and cannot be exhaustive. It is written by only a small selection of his large network of collaborators. Nevertheless we hope this will provide a useful summary for readers wanting to know more about the origins of work, events and software that are so widely relied upon by scientists today, and undoubtedly will continue to be so in the foreseeable future.