Terminology and Language Issues

Terminology Problems

In different models of the same species, parts with similar names may only denote roughly the same tissues.

In GALEN:	Lobe of left lung
Maps in FMA to:	Upper Lobe of left lung
	Lower lobe of left lung

More of a problem is that in anatomical models of different species, parts with the same (or similar) names do not always denote homologous tissues:

In Mouse “tail”	(posterior extension of the spine)
In C.Elegans “tail”	(male sex organ)

The model species we are comparing have many anatomical parts. For
example, Mouse has 3559 anatomical parts, Drosophila has 506
anatomical parts, and C. Elegans, 242 anatomical parts.

Does the language used in terminologies and anatomical ontologies suggest what parts may be similar? Yes. But not directly - context is crucial. The entire series of names from root to leaf node is needed to ground terms that are themselves underspecified. Grounding refers to connecting natural language expressions such as mouse tail, with a model of the world. In this case, the model is a mouse ontology.

Context is the key to clarifying terms such as "tail". If you look at the paths for tail in Mouse and C.Elegans, its clear that they are different:

Path name in Mouse:

- embryo

- tail

- nervous system

- peripheral nervous system

- segmental spinal nerve

Path name in C.Elegans:

- organ_system

- sex-associated system

- male-associated system

- male tail

- hypodermis

- fan hypodermis

Language processing techniques are used to help compare anatomical part names across species. They include:

a.	Normalizing terms to limit the effect of different descriptive styles.
b.	Comparing content words by removing stop words.
c.	Ensuring comparable forms of words by stemming and lemmatizing.
d.	Results are then treated as an unordered set.

The results of the language processing experiments show that the lexical suggestions were indicative of structural support in 75% of the comparisons.