Challenge LLL Syntactic Analysis Guidelines
Sophie Aubin 8th March 2005
1 Introduction
This document aims at describing the syntactic relations provided to the Challenge participants.
The syntactic parsing of the sentence is performed by the Link Parser1, version 4.1. The Link Parser data (grammar and dictionaries) were adapted to biological texts at MIG lab. The analysis product ed by the Link Parser being very rich and hardly understandable by non-specialists, it is filtered, interpreted and written in a more readable format. We created a set of syntactic relations that try to satisfy the following requirements :
• homogeneity
• expressivity
• conciseness
Section 2 describes the format we chose to express the different information given by the relations. Section 3 is an alphabetical list of the syntactic relations along with their definition and an example. Some help about english grammar can be found on http://www.ucl.ac.uk/internet-grammar/home.htm.
2 Relations format
Two major kinds of information are contained in a syntactic relation :
• the function of the relation that links 2 words (e.g. "subject")
• the morpho-syntactic nature of the 2 words linked by the relation (e.g. a "noun" and a "verb")
We then separated our relations names into 2 fields distinguished by the mark ":" that correspond to two different features :
function:nature
TDavy TEMPER.LEY. An Introduction to the Link Grammar Parser Carnegie Mellon University, March 1999. lit tp: / / www .link .cs. emu .edu/link/
1
The "function" feature is necessary while the "nature" feature is not. The "function" feature always starts with a generic function that can be specified with an attribute, separated from the generic name by an underscore ("_"). The specification can be either particular function (see "MOO All") or position (see "MOD_POST") or the preposition that participates in a ternary relation, e.g. for a complement introduced by the preposition "in", the function is "COMPJn".
The function feature is formed as follows (the square brackets mean that the element is optional) :
function [.specification]
The "nature" feature is composed of 2 elements corresponding to an artificial morpho-syntactic category (MSG) of the 2 words linked by the relation, separated by an hyphen ("-"). The first element is always the HEAD2 of the relation, the second being the EXPANSION3. The MSG is not directly issued from the analysis and does not necessarily correspond to the real MSG of the words. For instance, the second element of a SUBJ:V-N relation can be a pronoun. This part of the relation name is more or less informative depending on the kind of the relation (more in a MOD_ATT:N-N or MOD:V-ADV relation, less in a OBJ:V-N where the elements are always a verb and a noun or pronoun).
MSChead-MSCexpansion
There are 5 different morpho-syntactic categories :
• V : verb,
• V_PASS : passive verb,
• N : noun,
• ADJ : adjective,
• ADV : adverb
The table 1 in section 4 shows the relations that can occur between pairs of syntactic elements.
3   List of syntactic relations
Note that in the examples below, the words appear in the relations for sake of readability, as opposed to the challenge dataset format.4.
APPOS :
Apposition generally occurs between two nouns. This relation, however, does not give any information about the morpho-syntactic categories of the two elements involved.
2The HEAD is the governor of the relation, e.g. the verb in a verb-subject relation or the noun in an noun-adjective relation
3The EXPANSION is the governed element of the relation, e.g. the noun in a verb-subject relation or the adjective in a noun-adjective relation
4The FORMAT is syntactic_relation(wordl,word2)(id_wordl,id_word2) instead of rela-tion('syntactic_relation',id_wordl,id_word2)
2
A sigmaW dependent promoter ( PW ) precedes sigW. demonstrating that this transcription factor is positively autoregulated. APPOS(promoter,PW)
COMP_prep:ADJ-N :
Prepositional complement between an adjective and a noun. Dephosphorylation of SpoIIAA-P by SpoIIE is strictly dependent on the presence of the bivalent metal ions Mn2+ or Mg2+. COMP_on: AD J-N(dependent, presence)
COMP_prep:ADV-N :
Prepositional complement between an adverb and a noun Primer extension experiments and Northern blot analysis show that an active sigmaA dependent promoter precedes kdgR and transcription is terminated at the putative p independent terminator downstream of kdgT. COMP_of:ADV-N(downstream,kdgT)
COMP_prep:N-N :
Prepositional complement between a noun and a noun Dephosphorylation of SpoIIAA-P by SpoIIE is strictly dependent on the presence of the bivalent metal ions Mn2+ or Mg2+. COMP_of:N-N(presence,Mn2+)
COMP_prep:N-V :
Prepositional complement between a noun and a verb Evidence based on the use of modified and mutant forms of the phosphatase protein indicates that SpoIIE blocks the capacity of unphospho-rylated SpoIIAA to activate sigmaF until formation of the polar septum is completed.
COMP_to:N-V(capacity,activate)
COMP_prep:N-V_PASS :
Prepositional complement between a noun and a passive verb
The amino domain retains ability to be phosphorylated by the phos-
phorelay.
COMPlto:N-V_PASS(ability,phosphorylated)
COMP_prep:V-N :
Prepositional complement between a verb and a noun
These results suggest that YfhP may act as a negative regulator for the
transcription of yfh.Q, yfh.R, sspE and yfhP.
COMP_as:V-N (acts,regulator)
COMP_prep:V-V :
Prepositional complement between a verb and another verb These results demonstrate that sigmaK dependent transcription of gerE initiates a negative feedback loop in which GerE acts as a repressor to limit production of sigmaK. COMP_to:V-V (acts,limit)
COMP_prep:V-V_PASS :
Prepositional complement between a verb and a passive verb. This relation does not occur in the results because the head verb (almost always
3
a verb of perception like "seem") is informative in the present task. The product of the codY gene proved to be required for this repression. COMP_to:V-V_PASS (proved,required)
COMP_prep:V_PASS-N :
Prepositional complement between a passive verbe and a noun.
Northern blot and primer extension analyses indicated that yfhS is transcribed by E sigma E during sporulation. COMP_by:V_PASS-N(transcribed,E sigma E)
COMP_prep:V_PASS-V :
Prepositional complement between a passive verb and a verb Selective 2H-labeling. l'SG-labeling and isotopic heterodimers were used to distinguish contacts between and within monomers of the dimeric protein.
COMP_to:V_PASS-V (used,distinguish)
COMP_prep:V_PASS-V_PASS :
Prepositional complement between a passive verb and another passive verb. This relation does not occur in the results because the head verb (almost always a verb of opinion or perception like "believe" or "appear") is informative in the present task.
The signal peptide was considered to be consisted of 38 amino acids. COMP_to:V_PASS-V_PASS (considered,consisted)
MOD:ADJ-ADV :
Modifier between an adjective and an adverb
Dephosphorylation of SpoIIAA-P by SpoIIE is strictly dependent on the presence of the bivalent metal ions Mn2+ or Mg2+. MOD:ADJ-ADV(dependent,strictly)
MOD:ADJ-N :
Modifier between an adjective and a noun
sspG transcription also requires the DNA binding protein GerE. MOD:ADJ-N(binding,DNA)
MOD:ADV-ADV :
Modifier between an adverb and an other adverb
R factors indicate the structures fit the experimental NOE data very well. MOD:ADV-ADV(well,very)
MOD:V-ADV :
Modifier between a verb and an adverb
From these results we conclude that ComK negatively regulates degR expression by preventing sigmaD-driven transcription of degR. possibly through interaction with the control region. MOD: V-ADV(regulates,negatively)
MOD:V PASS-ADV :
Modifier between a passive verb and an adverb
In addition to the typical sigmaB dependent, stress- and starvation inducible pattern. yvyD is also induced in response to amino acid depletion.
MOD:V_PASS-ADV(induced,also)
4
MOD ATT: N-A D.I :
Attributive modifier between a noun and an adjective Transcription of ydhD was dependent on SigE, and the mRNA was detectable from 2 h after the cessation of logarithmic growth ( T2 of sporulation ).
MOD_ATT:N-ADJ(growth,logarithmic)
MOD ATT:N-N :
Modifier between a noun and a noun
Northern blot and primer extension analyses indicated that yfhS is transcribed by E sigma E during sporulation. M OD _ATT: N-N (analyses, extension)
MOD POST: N-A D.I :
Post-posed modifier between a noun and an adjective
It binds to a palindromic sequence, very similar to an Escherichia coli
Crp binding site, located upstream from arc A.
MOD_POST:N-ADJ(sequence,similar)
MOD PR ED:N-AD.I :
Predicative modifier between a noun and an adjective
Therefore, the physiological role of sigmaB dependent katX expression
remains obscure.
MOD_PRED :N- AD J (role,obscure)
MOD PR ED:N-N :
Predicative modifier between a noun and another noun
In that respect sigB is similar to the previously described gene gsiB which
is also a member of the sigmaB regulon.
MOD_PRED:N-N(gsiB,member)
NEG :
Negation between any morpho-syntactic category and the negation word. The morpho-syntactic categories of the two elements involved are not specified in the relation.
DNase I footprinting showed that SpoIIID binds strongly to two sites in the cotC promoter region, binds weakly to one site in the cotX promoter, and does not bind specifically to cotB. NEG(bind,not)
OBJ:V-N :
Object between a verb and a noun
DnaK, a general regulator of the heat shock response, which in bacteria inhibits the heat shock sigma factor sigma 32. OBJ:V-N(inhibits,factor)
OBJ:V_PASS-N :
Object between a passive verb and a noun
Transcription of the cotB, cotC, and cotX genes by final sigma(K) RNA polymerase is activated by a small, DNA-binding protein called GerE. OBJ:V_PASS-N(called,GerE)
5
SUBJ:V-N :
Subject between a verb and a noun
Expression of the sigma(K)-dependent cwlH gene depended on gerE. SUB J: V-N(depended,Expression)
SUBJ:V_PASS-N :
Subject between a passive verb and a noun. This stands for the patient of a passive verb.
Northern blot and primer extension analyses indicated that yfhS is transcribed by E sigma E during sporulation. SUBJ:V_PASS-N(transcribed,yfhS)
4   Relations between different kinds of Heads and Expansions
Not any relation can be found between two syntactic elements. The table 1 is a synthesis of the possible associations of different syntactic elements through relations. The vertical entries stand for the head of a relation and the horizontal entries stand for the expansion of a relation. For instance, an OBJ:V-N relation (object relation between a verb (head) and a noun (expansion)) is to be found on line 1 (V), column 3 (N).
Exp. Head	V	V-PASS	N	ADJ	ADV
V	COMP-prep	COMP_prep	COMP_prep SUBJ OBJ		MOD
V_PASS	COMP_prep	COMP_prep	COMP_prep SUBJ OBJ		MOD
N	COMP-prep	COMP_prep	COMP_prep MOD-ATT MOD-PR.ED	MOD-ATT MOD-POST MOD-PRED	
ADJ			MOD COMP_prep		MOD
ADV			COMP_prep		MOD
Table 1: Relations between different elements
6