Considering Bilingual Dictionaries Against a Corpus . Do English-French Dictionaries Present " Real English " ?

This article investigates the extent to which four representatives of the latest generation of English-French I French-English dictionaries present "real English", i.e. actually used meanings of actually used English word patterns. The findings of a corpus study of the verb CO!,!"SIDER are confronted with the entries for this verb in the English-French sections of these dictionaries, leading to the conclusion that there are important gaps in both the semantics and the lexicogrammar they cover, and that the organization of entries does not match the corpus frequency data. Corpus research could help to fill the gaps and should therefore be taken seriously by compilers of bilingual dictionaries.

corpus research.To be able to compete with HarperCollins' pioneering Collins COBUILD English umguage Dictionary, publishers like Cambridge University Press, Longman, and Oxford University Press have all produced learners' dictionaries whose covers boast that they are based on computerized COrpora ranging from 30 plus to 100 million words.The Collins Cobuild 1995 catalogue, announcing the publication of the second edition of the COBUILD dictionary, now proudly introduces the "Bank of English", "a data bank containing over 200 million words of modem English from hundreds of different sources".We clearly have the technology now to "help learners with real English", i.e. to present them with actually used meanings of actually used lexicogrammar, rather than with the perhaps mistaken intuitions of the arm-chair lexicographer, and not to use this technology would be inexcusable.
The same must be true for the bilingual dictionary, the language learners' tool par excellence, but here things seem to be developing more slowly.Of the three very recent English-French dictionaries dealt with in this article (Collins Robert 1993, Hachette Oxford 1994, and Larousse 1993), only one devotes a whole paragraph of the preface to corpus research: This is the first French and English dictionary to have been written using electronic corpora: two huge databases of electronic texts, one of current French and the other of current English.Each database contained over 10 million words of language in use.Access to these databases has provided accounts of words and their translations which are always authentic and often revealing.Users of the dictionary can feel confident that translations presented derive from study of real language as it has actually been used in a wide range of contexts.The resulting text provides modem idiomatic coverage of general French and English, with many new words, extensive treatment of colloquial expressions, and thousands of example sentences showing real language in action.(Hachette Oxford 1994: v) Another one is rather vague about a corpus in one single subclause: Historical events, breakthroughs in science and technology, new fashions in music, in eating habits and leisure activities all produce their own new vocabulary at a startling rate.Some of these new words, like the phenomena themselves, will fade and be forgotten.Others will last and become established.A good dictionary reflects such linguistic developments and our team of British and French lexicographers has been monitoring developments as closely as ever, aided by Collins' growing databank of real written and spoken language.(Collins Robert 1993: xiii) Moreover, these two paragraphs seem to suggest that the corpus only served to help in the decisions about which entries the dictionary should have and played no part in the construction of entries, i.e. in the decisions about which - ----------------------------------------------------Oleanings and which lexigrammatical patterns should be represented.
The third dictionary makes no mention of a corpus at all: Prepared over a period of several years by an international team of lexicographers based in London, Edinburgh and Paris, with contributions from consultants in several countries, this is the latest addition to a longstanding Larousse tradition of excellence in dictionary publishing.
[ ... ] Every effort has been made to ensure that example sentences reflect authentic usage and that translations are accurate and up to date.Detailed glosses are provided wherever it is necessary to clarify usage or avoid confusion.(Larousse 1993: vii) It may be, of course, that the last publisher is just less in touch with the latest trends in dictionary advertising in England than the others, and it is certainly not our intention to take the statements in these prefaces (or the lack of them) at their face value.Instead, what we propose to do in this article is to confront the three dictionaries already mentioned, as well as Larousse (1994), an abridged version of Larousse (1993) (because it might be interesting to compare the coverage of an abridged and an unabridged version), and the older, authoritative, Harrap's Standard Dictionary (Harrap 1980), with corpus data, with the aim to find out how well they represent "real English".2 The tester we are going to use is the verb CONSIDER.We will investigate its use in the LOB corpus, categorizing its meanings and the patterns it enters into, and establishing links between meanings and patterns.These findings will then be put next to the treatment of this verb in the English-French sections of the dictionaries under consideration.

2.
CONSIDER in the LOB corpus 3 The lexical item CONSIDER occurs 368 times in the LOB corpus; that is to say, there are 131 tokens of the base form consider, 42 tokens of the preterite form considered, 135 tokens of the past participle considered, 15 tokens of the s-form considers, and 45 tokens of the ing-form considering.The follOWing meanings can be distinguished: 4 2 3 4 An added investigation of how well they represent "real French" would make the article too bulky.The question is addressed in Noel, Defrancq and Devos (in prep.).The Lancaster-Oslo/Bergen (LOB) Corpus was compiled in the 19705 under the direction of Geoffrey Leech, University of Lancaster, and Stig Johansson, University of Oslo.It is a 1million-word corpus containing SOD 2,OOO-word text samples selected from texts printed in Great Britain in 1%1.A full description of the corpus is given in Johansson et al. (1978).We used the tagged version available on the !CAME Collection of English Language COIpora CO-ROM, distributed by the Norwegian Computing Centre for the Humanities.
The examples provided hint at the collocational evidence that could be adduced for this 1.
"have the opinion that (someone or something is something)" This is the meaning most frequently expressed by CONSIDER and when it does it is very often (for precise frequencies see Table 1) complemented by a noun phrase plus another constituent, either an adjectival phrase (I), another noun phrase (2), a to-infinitive (3), or a prepositional phrase which is normally introduced by as (4), though beyond also occurs (5).Quite often as well when it has this meaning, CONSIDER is followed by a that-clause (6).The patterns exemplified in (7) and ( 8) are rarer.( 7) is an example of free direct speech and in (8) the complement is a WH-clause.
(1) She attributes to me the words "the worst country under socialism is better than the best country under capitalism", and then claims that the labour party will only flourish when it gets rid of people like myself who "consider freedom relatively unimportant."(LOB Press:editorial B09:13) (2) "You may consider him a man, I would say he was a mentally adolescent cad." (LOB P:Romance,love story P02:92) (3) The speed with which water could be heated was considered to be particularly important.(LOB E:Skills,hobbies E25:88) (4) We do not consider the resolutions adopted by the consultation as in principle incompatible with the above statement.(LOB D:Religion 017:36) (5) With by wisdom compare Job 28.12, where the quest for wisdom is considered beyond man's capacity.(LOB O:Religion 003:70) (6) Although the number of boilermakers who are out of work has been reduced steadily during the past year, the SOciety does not consider that the time is opportune to relax the ban so far as welders, riveters, burners and heaters are concerned.(LOB Press:editorial B18:38) There must also be a better staffing ratio -16 students to one teacher in a kitchen were too many, he considered.(LOB E:Skills,hobbies E31:21) (8) The Prime Minister has never given the slightest indication who he considers should follow him in office.(LOB B:Press:editorial B08:41) 2.
"to look at or call to mind" Under this heading we subsume cases like (9), where a physical entity is observed, and which are rather rare, as well as more frequently occurring cases like (to), where what is observed is less palpable.
partitioning of the conceptual area covered by CONSIDER Arguing our way through this evidence would put us too much off course, however.We agree with Moon (1987: 86) that "descriptions in dictionaries need to satisfy intuition in order to be satisfactory overall" and the examples will go a long way towards enabling readers to judge whether our classification of senses meets this criterion at least. (9) (10) Again noun phrase complements (23) are the most frequent additions, occasionally supplemented with an as-phrase (24), and there is one occurrence each of WH-(25) and whether-clauses (26).Significant, however, is that ing-clause complements ( 27) can only occur in this category.
(23) The tragic part is that so many don't bother to consider promotions and sometimes even regard special offers as a waste of time.(LOB E:Skills,hobbies E30:57) (24) Interviews with about half these men quickly convinced the panel that any approach at sixty-four -which had been considered as a possible interim stage -was unlikely to be profitable, and a decision was taken to plan a scheme of preparation for retirement suitable for men who had just reached the age of fifty.(LOB F:Popular lore F02:5) (25) The touchstone, for a man of mature years considering what to take up next, must always be [ ... ] (LOB F:Popular lore F21:30) (26) The Prime Minister disclosed in the Commons last night that he had considered early in the parliamentary recess whether to mobilise reserves necessary to bring the British army of the Rhine on to a war footing.

"to bear in mind"
Again the commonest addition is a noun phrase (2B), and there is one WH candidate (29).Notable as well, however, is that a that-clause can also be used ( 30).
Certainly his compatriots in the labour party would do well to grasp the fact that the onus is now very definitely on local authorities to consider all sections of the community of ratepayers when formulating their rent policies, instead of merely where political advantage may be gained or lost.(LOB Press:editorial B23:47) "America!"she breathed, and for a moment could not make herself consider how her delight must wound him, compunction killed by the realisation of a thousand dreams.(LOB N:Adventure,western N2B:72) Many well-informed Americans welcomed his rising against President Batista, and consider that he only turned towards Moscow when he was rebuffed during his visit to the U.S.A. in 1959, perhaps chiefly because the American companies with investments in Cuba disliked his proposals for land reform.(LOB G:Belle lettres,biog G76:64) "to discuss in a meeting" Also here noun phrase additions (31) are the most typical complements, sometimes accompanied by a with-phrase (32).WH-clauses are possible complements as well (33).
(31) Senator Robertson's committee has to pass Mr Weaver's nomination before it can be considered by the full Senate.(LOB A:Press:reportage A01:56) (32) With summit diplomacy revived in a spectacular fashion by the Vienna meeting with Mr Kruschev, the President considered with the Prime Minister the chances of a meeting of the big four soon.(LOB A:Press:reportage A2B:107) (33) Even so headmaster and school had a conscience, and we know that at one time the assistant masters formed a committee of their own to consider what could be done educationally for the town.(LOB F:Popular lore F28:4B) Table 1 displays how often each meaning is present in the corpus and also provides a frequency breakdown per complementation type.Approaching these findings from the formal end, we can summarize the facts of the use of CONSIDER in "real English" as follows, listing the kinds of complements it can take according to diminishing frequency: CONSIDER + a single noun phrase can mean "to look at or call to mind", "to think carefully about something", "to deal with in a text", "to contemplate a possibility", "to bear in mind" and "to discuss". ii.
CONSIDER + a noun phrase + an adjectival phrase, another noun phrase or a to-infinitive can only mean "to have the opinion that somebody or something is something". iii.
CONSIDER + a noun phrase + a prepositional phrase can mean "to have the opinion" (normally as), "to look at or call to mind" (as and other prepositions), "to contemplate a possibility" (as) and "to discuss" (with).iv.
CONSIDER + that-complement either means "to have the opinion" or "to bear in mind".v.
CONSIDER + WH-complement can have any of the meanings specified. vi.
CONSIDER + ing-complement can only mean "to contemplate a future action". vii.
CONSIDER + if / whether-complement can mean "to think carefully about something", "to deal with in a text", or "to contemplate a possibility".viii.CONSIDER without any complements can only mean "to think carefully about something".
It is clear that the patterns CONSIDER enters into are not equally distributed over all its meanings.When comparing the corpus data about CONSIDER with its treatment in the dictionaries under consideration here, there are therefore three questions that need to be answered: 1.
Are all the meanings of CONSIDER represented? 2.
Are all the patterns CONSIDER enters into represented?3.
Are the right patterns linked up with the right meanings?
In addition, the order in which meanings and forms are presented must be looked at.It would seem reasonable to expect that dictionaries would treat the most frequent meanings and forms first and marginal meanings and forms last, so as to convey the right impressions about what is typical and less typical.An additional question is therefore: 4.
Does the organization of the material in the CONSIDER entries reflect the above frequency data?

Which meanings?
To facilitate answering the first of the questions formulated above -which meanings are represented?-we list in Table 3 which meaning glosses each of the dictionaries provides, indicating how they correspond to the distinctions we have made in section 2.
The numbers between brackets indicate the order in which the different senses are presented; senses that bear the same number are conflated by the dictionary in question.
It is obvious from Table 3 that there is some difference of opinion about the concept .alarea covered by CONSIDER and the way this should be partitioned.The following observations can be made in this respect: 1.
None of the dictionaries recognizes what we have identified as the fourth most frequent meaning of CONSIDER: "to deal with in a text".What separates this sense from the others, and to our minds therefore warrants it to be differentiated from the others, is that it is metatextual in nature: it is used to express what a writer (or speaker) does in a text or part of a text.A few more examples: (34) Before I consider the other dishes I must come on to the crown of lamb, for it is a splendid dish deserving a splendid wine.(LOB E:Skills,hobbies E19:96) (35) The bulk of my discussion so far could be regarded as an attempt to stress the importance of performance iri action, but now I want to consider intention.(LOB G:Belle lettres,biog G63:25) (36) Tonight I will be considering some aspects of life in Zululand and change has been as violent here as elsewhere on the continent.(LOB J:Learned,scientjfic J22:7) (37) The effect on the data of using applicators of different design or endplate thickness will be considered further in a separate publication.(LOB J:Leamed,scientific Jl5:22) More than 10% of all tokens of CONSIDER in the LOB corpus have this metatextual meaning, and separating it from the "to think about" meaning, as we have done, therefore does not seem unwarranted, especially because this kind of CONSIDER allows translations like traiter, parler de, passer il, which other uses of the verb do not allow.

2.
All other meanings we have identified are covered by the Larousse dictionaries, but Collins Robert and Hachette Oxford do not mention the "to look at" meaning, and neither Collins Robert, Hachette Oxford nor Harrap's mention the "discuss" meaning.An explanation for the first difference might be that the meaning "to look at something in the outside world", as exemplified in ( 9), ( 11), and also (38), is not very common these days and that cases where the considered entity is less palpable or not at hand to be looked at, like ( 10) and (39-40) are treated on a par with the "think about" instances.
(38) He leaned away, considering her, his eyes teasing.(LOB P:Romance,love story P13:94) (39) If one considers as a Whole the work done in modem languages in Scottish schools during the last few years, there is no doubt that the most significant advance has been in the field of understanding by ear and speaking the foreign language.(LOB H:Miscellaneous H03:44) (40) Consider, for example, the meeting-house at Old Meeting, Norwich, or Swanland, east Yorkshire, or Tadley, Hants; or any of those whose appearance is preserved only in faded prints in the vestries of more modem churches [ ... J (LOB D:Religion D02:6) Collins Robert, Hachette Oxford, and Harrap's would probably also treat instances where two or mor@ people jointly consider a question, and therefore "discuss" it, as instances of the "think about" meaning.In other words, some dictionaries probably do not recognize some of the meanings we have distinguished because they conflate them with other senses.

3.
Other meanings are covered but not treated separately.For instance, the "think about" and "contemplate a possibility" meanings are treated together under the same sense gloss in Collins Robert, Hachette Oxford and the Larousse dictionaries, and are dealt with in the same part of the entry in Harrap's as well, which uses sense glosses rather sparingly, letting the examples speak for themselves as it were.

4.
All dictionaries, except Collins Robert, sometimes make finer distinctions than we have done.Hachette Oxford covers our fifth meaning ("to contemplate a possibility, future action") under two separate sense glosses ("give thought to, study" and "envisage, contemplate"), and the Larousse dictionaries do the same with our sixth meaning ("to bear in mind"), fine-tuning it into "bear in mind" and "show regard for".In Harrap's the examples found under 2b and 4c seem to be covered by our fifth meaning.
-To be able to decide which partitioning of the semantic extension of CON-SIDER is the best it is not enough to consider the dictates of the monolingual source language corpus.The requirements of the target language have to be borne in mind as well, for one might argue that if meaning differences in the source language do not result in different translations, there is really no need to draW attention to the semantic distinctions.Two questions must be answered, therefore: are the meaning distinctions that can be made in the source language well represented, so that the USers of the dictionary can easily trace the senses for which they require equivalent expressions in the target language?and if different meanings require different translations is this made sufficiently clear?
As far as a) is concerned the dictionaries with the least empty slots in Table 3 might be scoring better than the others.Harrap's really only fills two slots here (since there are only two glosses), and is therefore perhaps doing the worst job as far as helping the user with identifying meanings is concerned.The Larousse dictionaries fill all slots except one.Interestingly, the only difference (meaningwise) between the abridged and unabridged versions of the Larousse is that in the former "faces" are not mentioned as examples of things that can be "contemplated" (our sense 2), and "candidates" as topics that can be "discussed" (our sense 7).
With respect to b) it may be pointed out that: Collins Robert, which does not distinguish between our sense 3 ("think carefully about / study something") and sense 5 ("contemplate a possibility / future action"), does recognize that a sentence like I hild not con-sider°ed taking it with me needs another verb than the ones proposed in the list of possible translations at the start of its section of the entry, and uses the verb ENVISAGER in the translation of the example sentence, but had the example been part of a different section in the entry, it would have been easier to give ENVISAGER the status of target language equivalent, on a par with CONSIDERER and EXAMINER, and ENVISAGER would have been less hidden away.The Larousse dictionaries do not explicitly differentiate between thinking about problems and thinking about possibilities either, also proposing CONSIDERER and EXAMINER for both, but in the translation of the example sentence have you ever considered becoming an actress? the verb SONGER is used.Again, if this part of the entry had been split up, SONGER could have been given more prominence.The Larousse dictionaries split up our "bear in mind" meaning in "bear in mind points, facts, etc." and "show regard for feelings and wishes" proposing prendre en consideration and tenir compte de for the first meaning an'd only tenir compte de for the second, but an example of the second meaning, he has a wife and family to consider, receives the translation i/ a une femme et une famille a prendJ'e en considemtion, perhaps proving that making this meaning distinction is not really necessary.
In other words, the demands of the target language seem not to contradict the findings of our corpus study.Meaning differences in one language often result in different translations in another, which makes it all the more necessary that these differences in meaning receive due attention.

Which forms (for which meanings)?
All five dictionaries under consideration use explicit grammatical information rather sparingly.Collins Robert and Har~ap's inform us that CONSIDER is a transitive verb, the Larousse dictionaries also make mention of its intransitive use, and Hachette Oxford also explicitly mentions its use as a reflexive verb and the use of the form considered as an attributive adjective.None of the dictionaries, however, offers an exhaustive list of the kinds of complement CON-SIDER allows.Instead, they present a fairly fragmentary picture of its grammatical potential through the use of examples (see Table 2).One might argue that grammar belongs in explanatory learners' dictionaries, not in bilingual dictionaries, but this would be arguing the wrong way: if not all patterns go with all meanings, then the inclusion of grammatical information may be very useful in helping users to disambiguate words, facilitating the choice of an eqUivalent word in the target language.Users of Harrap's, for example, are at a disadvantage because the consider +ing-form pattern is not exemplified.How are they to decide on the meaning of consider in (41)?If they mistake Dr. Verwoerd for a scientist, they might interpret (41) to mean that he refuses to study the abolition of apartheid.
(41) They point out that Dr Verwoerd refuses to consider abandoning the apartheid policy.(LOB Piess:editorial B01:17) Supplying incomplete information may do more harm than good.Hachette Oxford, for instance, mentions the pattern to consider sb/sth as 5th under the gloss "envisage, contemplate" (our sense 5), but does not exemplify it under "regard" (our sense 1).This would lead to a correct interpretation of ( 42), where the proposed translation penseJ" aqn/qch come qch can be used, but to a misinterpretation of ( 43), where considerer comme is a better translation.
(42) Or, a Middlesex federation of labour parties working within a regional council covering the northern home counties might be considered as a possible solution.

Ranking senses (and forms)
Another glaring difference between the corpus data as we presented them and the dictionaries, and between individual dictionaries, is the order in which the different senses are presented.There is a fair amount of agreement between the ranking in the Larousse dictionaries and the corpus frequency data (the only deviation being the result of a difference in the partitioning of the semantic field covered by the verb), but there are greater differences with and between the other dictionaries.Most striking is that Harrap's starts with a use of CON-SIDER which it qualifies as "archaic" and that Collins Robert, Hachette Oxford, as well as Harrap's end with the meaning with which it is most frequently used.A ranking of both senses and forms more in line with frequency data would not only add to the userfriendliness of the dictionaries -because the more likely uses would be retrieved faster -it would also paint a truer picture of both the conceptual area covered by the entries as well as their typicallexicogrammatical patterning. .

Conclusion and a plea
When considering bilingual dictionaries against a corpus of real language (however modest in scope this little exercise has been), it becomes clear that there are.important gaps in both the semantics and the lexicogrammar they cover.It is interesting to note that the Larousse dictionaries, which do not make explicit mention of corpus research, stand comparison with the two recent ones that do: they do not contain more gaps, just different ones.Hachette Oxford, which is the most explicit about the use of a corpus, is slightly better in its coverage of grammatical patterns, but not in its coverage of meanings, and it does not structure entries on the basis of frequency of use.Collins Robert, whose reference to corpus research is hazier, is the least detailed of the newer dictionaries, offering fewer meanings and fewer patterns, and here as well the organization of entries does not seem, to be based on a frequency study.The older Harrap's dictionary presents many examples, but hardly any explicit semantic information, though entries seem on the whole to be organized according to meaning.An often-heard complaint from language teachers is that "students can't use dictionaries" -"they just pick the first translation the dictionary offers without thinking about its appropriateness" -but the fact is that dictionaries could do more to make it easier for them to make judgements about appropriateness.All the dictionaries we have examined here already do a lot in this respect, in different ways and to varying degrees, but there is certainly room for improvement.What is needed is, first of all, greater explicitness about the meanings of entry words and about the relation between meaning and form 5 , and second, an organization of entries on the basis of frequency.Corpus research could do much to help to fill the gaps and is a sine qua non to meet the second need.Compilers of bilingual dictionaries would therefore be well-advised to take corpora seriously.(3) bear in mind: (3) bear in mind: points, facts; costs, points, facts: costs, difficulties, dangers difficultie., dangers (4) show regard for: (4) show regard for: feelings, wishes feeling., wishes (5) discuss: report, (5) discuss: report,

Table 2
displays the entries for CONSIDER in the five dictionaries we have examined, not in the way they can be found there, but cast into the mould of Reproduced by Sabinet Gateway under licence granted by the Publisher (dated 2011) http://lexikos.journals.ac.zaConsidering Bilingual Dictionaries Against a Corpus 65 The importance has been previously stressed of considering the production, handling, storage, packaging and processing of food as links in one continuous chain of operations, the final objective of which is to provide the nation with food of the highest quality at the lowest economic price.(LOBHMiscellaneousHlO:32)Othernoteworthy gaps are the absence of the CONSIDER + that-clause construction in the Larousse dictionaries, the CONSIDER + NP + to-infinitive construction in Collins Robert, the intransitive use in Collins Robert.Worth mentioning as well is that most dictionaries supply very little grammatical information, either explicitly or in the form of examples, about the CONSIDER + NP pattern, unless one treats as such the nouns which form part of the sense glosses, like (think about) problem, possibility in Collins Robert, (take into account, bear in mind) {risk, cost, difficulty, malted in Hachette Oxford, and {contemplate -face, picture, scene] in Larousse.

Table 2 :
CONSIDER in English / French dictionaries