POOLING AND THREADING IN TRANSLATION TERM BANKS 
Merle D. Tenney 
Automated Language Process. Systems, 750 
North 200 West. 
Pro, o, Utah 84601 USA 
Robert A. Russell 
Weldner Communications, Inc. 
1673 West 820 North, 
l~OVO, Utah 84601, USA 
Bilingual end multillngual terminology data banks 
(term banks) have proven an effective aid to translation. 
The translation problem which has been most resistant 
tO term bank assistance~ however, is that of identifying 
precisely those translations which are appropriate for a 
given term in a SL text. This is compounded by the related 
problems of SL homonymy and polysemy and TL synonymy. The 
greatest challenge at the text level is dynamically specify- 
ing the lexicon used in the subl~age of a document. The 
solution to this problem depends somewhat on whether termino- 
logy is held in one large bank, whic h promotes sharing and 
comprehensiveness, or in many smaller banks, whiohencoura - 
ges independence and specialization. 
The principle approach to the translation identificat- 
ion problem heretofore has consisted in filtering out ir- 
relevant translations on the basis of such accompanying do- 
cumentation as subject field, source indication, or grammat- 
ical code. Some organizations have developed initial capabi- 
lities for pooling or threading in order to deal with the 
sublanguage problem. ~ refers to the creation of inter- 
nally cohesive subinventories of a data base. Threadin~ 
refers to the ability to specify (and access) a logical file 
made up of a sequence of pools, in wb_ic~ the logical keys 
are associated with physical records on a flrst-hit basis. 
- 287 - 
OuT paper further elaborates those concepts and discusses 
their utility in term banks designed to support translation. 
Pools may be organized on any functional principle. They 
may reflect the admAnistrating body - from standards or~anAzat- 
ion through oompan~ divis¢on to individual translator. Or they 
may reflect a domain of application - from technical field 
through company product line to chapter in operations manual. 
3oms pools are created to override other more general pools, 
such as SL or TL dialect pools which preempt standard..langus~e 
pools. 
The object of threadin~ is the dynsunic creation of logic- 
al files speoift"cally tailored to ~he applications at hand° 
Y~aportant features of threading include the ability to access 
a select battery of pools~ the ability to prioritizs these 
pools by their task relevance, ,typically in a sequence from 
most specific to most general, and the ability, in interactive 
applications, to modify logical file records by granting read/ 
write access to one or more leading pools in the operative 
sequence° Application programs utilize one of 'three for~ of 
aocessz s£n~e access, which retrieves data solely from the 
first sequenced pool in which a term appears! multiple access, 
which makes data available from any of the sequenced pools in 
whiOh a term appears! and composite access, a special case of 
multiple access which constructs a data complex from the full 
set of sequenced pools in which a term appears.. 
The principal motivation for pooling and threating lles 
in their use in direct aids to translation - text-related 
g~ossary generation and text editing with term look-up and 
other interactive aids° Tn addition, they extend the capabilit- 
ies of maAntenanoe programs, such as glossary publication and 
foreign language instruction. 
They are some drawbacks to the use of pooling and thre- 
Mings as pools are proliferated, the potential for redundant 
entries increases! the problems inherent in relating files of 
- 288 - 
distinct types are cempounded; and the need for taxonomiee, 
selectional codes and documentation , and heuristic mechanisms, 
though mitigated, is retained. On the other hand, use of the- 
se devices yields improvements in the identification of per- 
tinent translations, in the flexibility of logical file con- 
struction, in the adaptability of available term banks, in 
the organization and management of termihological data, in 
the size requirements fordiversified term banks, in the 
portability of existiug terminology resources, and in ~he con- 
comitant potential for added revenue to terminology holders. 

References

Brinkman, Karl-Heinz. 1981. ,,Machine Aids to Translation." 
METAi Journal des traduoteur~/Tra~slator°s Journal, 
26, 67-75. 

Goetschalckx, Jacques. 1979..EURODICAUTOM." Translatin~ and 
the Computer, ed. Barbara M. Snell, 71-75. Amsterdam= 
North-Holland= 

Kittredge, Richard. 1978. "Textual Cohesion within Sublangua- 
ges= Implications for Automatic Analysis and Synthesis." 
Paper presented at the 7th International Conference on 
Computational Linguistics, Bergen, Norway, 14-18, Aug. 
1978. 

Masterman, Margaret. 1979, "The Essential Mechanism of Machi- 
ne Translation." Paper read to the British Computer 
Society, Jan. 1979. 

Schultz, Joachim. 1980. "A Terminology Data Bank for Trans- 
lators (TEAM)." META! Journal des traducteurs/Trans- 
lat0r'e Journal., 25, 211-29. 

Smith, Raoul N. 1978. "Computational Bilingual Lexicography: 
A la recherche du mot ~uste." Computer Support to 
Translation, ed. Jim Mathias, 84-112. Washington, D.C.s 
Foreign Broadcast Information Service. 

Z6msta, Ladlslav. 1973° "The Shape of the Dictionary for 
Mechanical Translation Purposes." IBsues in Linguistics: 
Papers in Honor of Henr~ and Ren~e Kahane, ed. BraJ B. 
Kaohru, Robert B. Lees, Yakov Malkiel, Angellna Pietran- 
gelll, and Sol Saporta, 912-22. Urbana, Ill.: Universi- 
ty of Illinois Press. 
