VM-HyprLex: Appl 2d

VM-HyprLex: Application 2d

VERBMOBIL Semantic Database V 0.5.3

Johannes Heinecke, Karsten Worm, 10 January 1996
Information on access

The present form of the Semantic Database has been provided with a new access key set (select `Key'):
  1. First attribute string identity
  2. First attribute substring
  3. Global substring match over the whole database
  4. Attribute-Value conjunction
The option `SemDB' permits generation of subdatabases with of the whole entry set, but only marked attributes. The `All' option generates the whole database. For further information consult the general HyprLex FAQ and context-sensitve information sources.

Attribute names, SemDB descriptors and descriptions

  1. LeX4-Base: LeX4-Base
  2. SemLemma: Lemma
  3. IMS-POS: POS
  4. PredName: PredName(s)
  5. SemClass: SemClass
  6. PredScheme: PredScheme(s)
  7. SemSort: Sorts
  8. Comments: {Comments}
  9. SemPOE: POE
  10. Designator: Designator
Source

The database dbsem.liste-0.5.3 for this application was provided by Johannes Heinecke (HU Berlin) and Karsten Worm (U Saarbrücken) as announced at the VERBMOBIL Lexical Semantics Workshop, Berlin (29.11-1.1295).

The internal database format was defined at the Semantics group meeting, Saarbrücken (13.-14.11.95) as follows:

Lemma POS PredName(s) SemClass PredScheme(s) Sorts {Comments} POE


and extended in the current version to the following:

Lemma POS PredName(s) SemClass PredScheme(s) Sorts {Comments} POE


The attribute names used in the VM-HyprLex integration are shown in the attribute display selection menu and the table above.

Note on the two-level lemma concept (D. Gibbon, revised 13 Jan 95)

Criteria for morphological (inflectional) and semantic (conceptual) lemma format ion are not identical, therefore the MorLemma set (based on phonological and word-prosodic paradigms) and the SemLemma set are not co-extensive. For this reason, clarified by discussions with Martin Emele, I have introduced a new two-level lemma concept for the VERBMOBIL LexDB:
  1. MorLemmata in LexDB which are not string-identical to any SemLemmata in SemDB
  2. SemLemmata in SemDB which are not string-identical to any MorLemmata in LexDB
  3. String-identical MorLemmata in LexDB and SemLemmata in SemDB
The relation between SemLemma and MorLemma sets must still be defined, and minor revision of the MorLemma set will be necessary. When the SemLemma-MorLemma relation has been defined (as a table), assignment to inflected forms is straightforward. The LeX4-Base attribute represents the internal base lemma for the orthographic morphology in the underlying SemDB database (Berlin), and introduces the two-level lemma concept to the orthographic domain. The mapping between the LeX4-Base and MorLemma attributes is still to be defined.
VM-HyprLex service 10.12.95