;;; Copyright (c) 1991--2017 ;;; John Carroll, Ann Copestake, Robert Malouf, Stephan Oepen, Ben Waldron; ;;; see `LICENSE' for conditions. ;;; bmw (aug-03) ;;; - fixed code broken by *lexicon*-related changes ;;; aac (aug-03) ;;; - moved old flat file stuff to slex.lsp - added to cvs ;;; but not loaded by lkb.system ;;; - removed `other' entries from *lexicon* ;;; (code for handling them is either here or in ;;; specific files) (in-package :lkb) ;;; Lexical entries, storage etc ;;; Lexical entries are indexed by orthography (string) ;;; They are also all indexed by an identifier ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (defparameter *lex-database-default-extra-mode* :union) (defvar mrs::*initial-semantics-path*) (defvar *lexicon-in*) (defvar *verbose-lex-lookup-word* nil) (defvar *lexdb*) (defvar *postgres-mwe-enable*) (defvar *lexicon-lexical-entries-cache-p* t) ;; remove this? (defclass lex-database () ((lexical-entries :initform (make-hash-table :test #'equal)) ;;; cache results of lookup-word ;;; :empty records cases where lookup-word returned nil ;;; (to distinguish from cases where value not cached) (psorts :initform (make-hash-table :test #'eq)) ;;; cache results of read-psort ;;; :empty records cases where read-psort returned nil ;;; (to distinguish from cases where value not cached) (extra-lexicons :initform nil :accessor extra-lexicons) (extra-mode :initform *lex-database-default-extra-mode* :accessor extra-mode) (part-of :initform nil :accessor part-of) (invalid-p :initform t :accessor invalid-p) (cache-lex-list :initform nil :accessor cache-lex-list) (name :initform nil :accessor name) )) (defmethod link ((sub-lexicon lex-database) (lexicon lex-database)) (if (eq sub-lexicon lexicon) (error "cannot link a lexicon to itself!")) (with-slots (extra-lexicons) lexicon (with-slots (part-of) sub-lexicon (unless (and (member sub-lexicon extra-lexicons) (member lexicon part-of)) (push lexicon part-of) (push sub-lexicon extra-lexicons)))) lexicon) (defmethod unlink ((sub-lexicon lex-database) (lexicon lex-database)) (with-slots (extra-lexicons) lexicon (with-slots (part-of) sub-lexicon (setf part-of (remove lexicon part-of)) (setf extra-lexicons (remove sub-lexicon extra-lexicons)))) lexicon) ;;; ;;; given the orthography (string in all upper case), look up all lexical ;;; entries that _contain_ that orthography; return list of identifiers. ;;; (defgeneric lookup-word (lexicon orth &key (cache))) ;;; ;;; for a given lexicon, return all words, i.e. strings of orthography that ;;; occur as part of the spelling of a lexical item; returns "ad" and "hoc" as ;;; two separate `words' but not "ad hoc". ;;; (defgeneric lex-words (lexicon)) (defgeneric lexicon-loaded-p (lexicon)) (defgeneric read-cached-lex (lexicon filenames)) (defgeneric store-cached-lex (lexicon)) (defgeneric set-lexical-entry (lexicon orth id new-entry &key cache)) (defgeneric close-lex (lexicon &key in-isolation delete)) (defgeneric collect-expanded-lex-ids (lexicon)) (defgeneric store-psort (lexicon id entry &optional orth)) (defgeneric read-psort (lexicon id &key cache recurse new-instance)) (defgeneric unexpand-psort (lexicon id)) ;;; ;;; for a given lexicon, return list of psort identifiers (as strings), i.e. ;;; identifiers of lexical items plus start symbols and node labels (for the ;;; time being, these are stored in the same namespace). ;;; (defgeneric collect-psort-ids (lexicon &key cache recurse)) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; This isn't getting used correctly, since it's not preserved in the ;; lexical cache. But it is only used for outputting source ;; files in other syntaxes and batch checking, where ;; it is acceptable to have to read in a fresh lexicon (defvar *ordered-lex-list* nil) (defparameter *batch-mode* nil "set when indexing to prevent errors in expanding a lexical entry being signalled as continuable errors, rather than written to a file.") (defstruct (psort) id unifs def-unifs full-fs language comment) (defstruct (lex-entry (:include psort)) orth infl-pos ; for a multi-word entry only - a number ; indicating the element that can be inflected ; or NIL - no inflection ) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Database-independent lexical access functions ;;; (defun get-lex-entry-from-id (id &key (cache t)) ;;; now lexical entries only (let ((entry (get-unexpanded-psort-entry id :cache cache))) ;; if we haven't previously expanded then ;; destructively modify entry (when entry (cond ((eql (lex-entry-full-fs entry) :fail) nil) ((lex-entry-full-fs entry) entry) ((expand-psort-entry entry) entry) (t (setf (lex-entry-full-fs entry) :fail) nil))))) ;; (bmw) copied from above (defun get-expanded-lex-entry (entry) (when entry (cond ((eql (lex-entry-full-fs entry) :fail) nil) ((lex-entry-full-fs entry) entry) ((expand-psort-entry entry) entry) (t (setf (lex-entry-full-fs entry) :fail) nil)))) (defun get-unexpanded-psort-entry (id &key (cache t)) ;; for multi words, where we don't want the full-fs to be created ;; until we're sure we've got all the bits (let ((entry (read-psort *lexicon* id :cache cache))) entry)) (defun clear-expanded-lex nil (empty-cache *lexicon*)) (defun clear-non-parents nil (format t "~%Removing cached lexical entries") (clear-expanded-lex)) (defun get-psort-value (psort) (let ((psort-entry (get-lex-entry-from-id psort))) (if psort-entry (psort-full-fs psort-entry) (let ((other-entry (get-other-entry psort))) (if other-entry (psort-full-fs other-entry) (let ((gr-entry (get-grammar-rule-entry psort))) (if gr-entry (psort-full-fs gr-entry) (let ((lr-entry (get-lex-rule-entry psort))) (if lr-entry (psort-full-fs lr-entry) (lex-expansion-error "Return nil" (format nil "~A is not a valid structure identifier" psort))))))))))) (defun get-lex-entry (orth) (loop for psort in (remove-duplicates (lookup-word *lexicon* orth)) for entry = (get-lex-entry-from-id psort) when entry collect entry)) (defun get-unexpanded-lex-entry (orth) (loop for psort in (remove-duplicates ;; is remove-duplicates necessary??? (lookup-word *lexicon* orth)) for entry = (get-unexpanded-psort-entry psort) when entry collect entry)) (defun generate-unknown-word-entries (orth) (loop for type in *unknown-word-types* nconc (let* ((*safe-not-to-copy-p* nil) (unifs (cons (make-unification :lhs (create-path-from-feature-list nil) :rhs (make-u-value :type type)) (make-unknown-word-sense-unifications orth))) (tdfs (let* ((fs (process-unifications unifs)) (wffs (if fs (create-wffs fs)))) (if wffs (construct-tdfs wffs nil t))))) (if tdfs (list (cons (gensym "unknown") tdfs)))))) (defun add-lex-from-file (orth sense-id fs-or-type defs) (let* ((lex-id (if orth (make-lex-id orth sense-id) sense-id)) (orth-string (if (and orth *sense-unif-fn*) (format nil "~A" orth) (extract-orth-from-unifs fs-or-type) )) (infl-pos (if (and (listp orth-string) (cdr orth-string)) ;; infl-pos is only relevant for multi-word entries (find-infl-pos fs-or-type orth-string sense-id)))) ;; adapted for the case where the orthography is only specified ;; in the FS; extract-orth-from-unifs must be defined on a ;; per-grammar basis (set-lexical-entry *lexicon-in* orth-string lex-id (make-lex-entry :orth orth-string :infl-pos infl-pos :id lex-id :unifs fs-or-type :def-unifs defs)))) (defun extract-orth-from-unifs (unifs) ;; returns a list of strings ;; doesn't do much error checking (let ((matching-unifs (loop for unif in unifs nconc (if (unification-p unif) (let ((unif-lhs (unification-lhs unif))) (if (general-path-match unif-lhs *orth-path*) (list unif))))))) (when matching-unifs (let ((n (length *orth-path*)) (orth-strings nil)) (cond ((null matching-unifs) nil) ((and (not (cdr matching-unifs)) (general-path-match (unification-lhs (car matching-unifs)) *orth-path* t)) (list (u-value-type (unification-rhs (car matching-unifs))))) (t (loop (let ((exact-match (dolist (unif matching-unifs) (when (let* ((path (unification-lhs unif)) (feats (path-typed-feature-list path))) (if (or (and (path-p path) (eq (nth n feats) (car *list-head*))) (and (typed-path-p path) (let ((nth-el (nth n feats))) (and nth-el (eq (type-feature-pair-feature nth-el) (car *list-head*)))))) (u-value-p (unification-rhs unif)))) (return unif))))) (unless exact-match (return)) (push (u-value-type (unification-rhs exact-match)) orth-strings) (incf n))) (nreverse orth-strings))))))) (defun general-path-match (unif-path path &optional exactp) (or (and (typed-path-p unif-path) (typed-path-matches (typed-path-typed-feature-list unif-path) path exactp)) (and (path-p unif-path) (path-matches (path-typed-feature-list unif-path) path exactp)))) (defun path-matches (tfplist flist exactp) (cond ((and (null flist) (null tfplist)) t) ((null flist) (not exactp)) ((null tfplist) nil) ((eq (car tfplist) (car flist)) (path-matches (cdr tfplist) (cdr flist) exactp)) (t nil))) (defun typed-path-matches (tfplist flist exactp) (cond ((and (null flist) (null tfplist)) t) ((null flist) (not exactp)) ((null tfplist) nil) ((eq (type-feature-pair-feature (car tfplist)) (car flist)) (typed-path-matches (cdr tfplist) (cdr flist) exactp)) (t nil))) (defun extract-orth-from-fs (tdfs) ;;; returns a single string, possibly concatenating the words (let ((fs (tdfs-indef tdfs)) (current-orth-path *orth-path*) (alt-orth-path *alt-orth-path*) (orth-strings nil)) (let ((simple-value (get-value-at-end-of fs current-orth-path))) (if (and simple-value (stringp simple-value)) simple-value (let ((current-orth (or (get-value-at-end-of fs (append current-orth-path *list-head*)) (let ((alt-value (get-value-at-end-of fs (append alt-orth-path *list-head*)))) (when alt-value (setf current-orth-path alt-orth-path)) alt-value)))) (loop (when (or (null current-orth) (not (stringp current-orth))) (return)) (setf current-orth-path (append current-orth-path *list-tail*)) (push current-orth orth-strings) (push " " orth-strings) (setf current-orth (get-value-at-end-of fs (append current-orth-path *list-head*)))) (apply #'concatenate 'string (nreverse (cdr orth-strings)))))))) (defun make-lex-id (orth sense-id) (intern (format nil "~A_~A" orth sense-id))) ;;; When expanding a lexical entry we want to eventually produce a ;;; tdfs which then has the non-persistent defaults incorporated. As ;;; an interim stage, we want a fs with all defaults which has not yet ;;; been linked. We also have the unifications formed from the ;;; orthography etc to unify in to the non-default fs. (defun lex-expansion-error (string1 string2) (if *batch-mode* (format t "~%~A" string2) (cerror string1 string2))) (defun lex-entry-local-fs (entry) (expand-psort-entry entry t)) (defun expand-psort-entry (entry &optional local-p) (let* ((*safe-not-to-copy-p* nil) (orth (lex-entry-orth entry)) (lex-id (lex-entry-id entry)) (language (lex-entry-language entry)) (fs (append (lex-entry-unifs entry) (when (and orth *sense-unif-fn*) (apply *sense-unif-fn* (list orth (format nil "~A" lex-id) language)))))) (process-unif-list lex-id fs (lex-entry-def-unifs entry) entry *description-persistence* local-p))) (defun make-non-lex-psort-entry (name constraint default) ;;; called for category templates, roots, idioms etc ;;; expansion happens at load time (let* ((*safe-not-to-copy-p* nil) (entry (make-psort :id name :unifs constraint :def-unifs default))) (process-unif-list (psort-id entry) (psort-unifs entry) (psort-def-unifs entry) entry *description-persistence*) entry)) (defun process-unif-list (lex-id indef-list default-specs entry persistence &optional local-p) (let* ((fs (process-unifications indef-list)) (indef (if fs (create-wffs fs)))) (if local-p fs (if indef (let* ((default-fss (loop for default-spec in default-specs collect (make-equivalent-persistence-defaults indef (car default-spec) (cdr default-spec) lex-id))) (local-tdfs (construct-tdfs indef default-fss t))) (let ((interim-fs local-tdfs)) (setf (tdfs-tail interim-fs) (yadu-general-merge-tails (tdfs-tail interim-fs) (tdfs-tail (ltype-tdfs (get-type-entry (type-of-fs indef)))) indef)) (let ((incorp-fs (if persistence (make-indefeasible interim-fs (list persistence)) interim-fs))) ;; Cut off useless pointers, to help garbage collection (compress-dag (tdfs-indef incorp-fs)) (setf (psort-full-fs entry) incorp-fs)))) (progn (if fs (format t "~%Structure for ~A could not be made well formed~%" lex-id) (format t "~%Structure for ~A could not be created~%" lex-id)) nil))))) ;;; Code for `other' entries ;;; utility function (defun get-other-entry (fs-id) (or (get-root-entry fs-id) (get-display-template-entry fs-id) (get-idiom-entry fs-id) (get-temporary-entry fs-id))) ;;; temporary storage of structures created by unification check (defparameter *temporary-entries* nil) (defun store-temporary-psort-entry (name fs) (push (cons name (make-psort :id name :full-fs fs)) *temporary-entries*)) (defun get-temporary-entry (id) (cdr (assoc id *temporary-entries*))) ;;; root stuff goes here, for want of anywhere better (defparameter *root-entries* nil) (defun clear-root-entries nil (setf *root-entries* nil)) (defun add-root-entry (id non-def defs) (push (cons id (make-non-lex-psort-entry id non-def defs)) *root-entries*)) (defun get-root-entry (id) (cdr (assoc id *root-entries*))) ;;; end code for `other' entries ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; General lexicon methods ;;; (defparameter *lookup-word-remove-duplicates* nil) (defmethod lookup-word :around ((lexicon lex-database) orth &key (cache *lexicon-lexical-entries-cache-p*)) (let* ((value (if (next-method-p) (call-next-method))) (mode (extra-mode lexicon)) (extra ;:if no vals in lexicon (or union mode) ... (when (or (null value) (eq mode :union)) (loop with result = nil for lexicon in (extra-lexicons lexicon) for value = (and lexicon (lookup-word lexicon orth :cache cache)) ;:shadow mode returns first set of vals found when (and value (eq mode :shadow)) do (return value) else when value do (setf result (append result value)) finally (return result)))) (value (append value extra))) value)) (defmethod lex-words :around ((lexicon lex-database)) (let* ((words (if (next-method-p) (call-next-method))) (extra (loop for lexicon in (extra-lexicons lexicon) for extra-words = (and lexicon (lex-words lexicon)) collect extra-words)) (word-lists (cons words extra))) (remove-duplicates (apply #'append word-lists) :test #'equal))) (defmethod collect-psort-ids :around ((lexicon lex-database) &key (cache t) (recurse t)) (let* ((ids (if (next-method-p) (call-next-method))) (extra (if recurse (loop for lexicon in (extra-lexicons lexicon) for extra-ids = (and lexicon (collect-psort-ids lexicon :cache cache :recurse recurse)) collect extra-ids))) (id-lists (cons ids extra))) (remove-duplicates (apply #'append id-lists) :test #'equal))) (defmethod set-lexical-entry ((lexicon lex-database) orth id new-entry &key (cache *lexicon-lexical-entries-cache-p*)) (store-psort lexicon id new-entry orth) (when cache (with-slots (lexical-entries) lexicon (dolist (orth-el orth) (pushnew id (gethash (string-upcase orth-el) lexical-entries)))))) ;;; todo: cache (defmethod read-psort :around ((lexicon lex-database) id &key (cache t) (recurse t) (new-instance nil)) (cond ((and (next-method-p) (call-next-method))) (recurse (some #'(lambda (lex) (read-psort lex id :cache cache :new-instance new-instance)) (extra-lexicons lexicon))))) (defmethod close-lex :around ((lexicon lex-database) &key in-isolation delete) (declare (ignore delete)) (with-slots (invalid-p extra-mode extra-lexicons part-of) lexicon (empty-cache lexicon) (setf invalid-p nil) (setf extra-mode *lex-database-default-extra-mode*) (if (next-method-p) (call-next-method)) (unless in-isolation ;;unlink from sub-lexicons (mapcar #'(lambda (lex) (unlink lex lexicon) ;;clear sub-lexicon (if (null (part-of lex)) (close-lex lex))) extra-lexicons) ;;unlink from super-lexicons (mapcar #'(lambda (lex) (unlink lexicon lex)) part-of)) lexicon)) (defmethod empty-cache :around ((lexicon lex-database) &key (recurse)) (with-slots (lexical-entries psorts cache-lex-list) lexicon (if (typep lexicon 'external-lex-database) (call-next-method)) (clrhash lexical-entries) (clrhash psorts) (setf cache-lex-list nil) (when (fboundp 'clear-lexicon-indices) (funcall (symbol-function 'clear-lexicon-indices))) (when recurse (mapcar #'(lambda (x) (empty-cache x :recurse recurse)) (extra-lexicons lexicon))) lexicon)) (defmethod compress-lexicon ((lexicon lex-database)) (maphash #'(lambda (id value) (declare (ignore id)) (when (and (lex-entry-p value) (lex-entry-full-fs value)) (compress-dag (tdfs-indef (lex-entry-full-fs value))))) (slot-value lexicon 'psorts))) ;;; End of general methods (defmethod collect-expanded-lex-ids ((lexicon lex-database)) (let ((ids nil)) (maphash #'(lambda (id value) (when (and value (not (eq value :empty)) (lex-entry-full-fs value)) (push id ids))) (slot-value lexicon 'psorts)) ids)) (defmethod lexicon-for-id ((lex lex-database) id) (if (read-psort lex id :recurse nil) lex (some #'(lambda (x) (lexicon-for-id x id)) (extra-lexicons lex)))) (defmethod forget-psort ((lexicon lex-database) id) "remove cached entry (can be :empty) from all lexicons" (remhash id (slot-value lexicon 'psorts)) (and (next-method-p) (call-next-method)) (some #'(lambda (x) (forget-psort x id)) (extra-lexicons lexicon))) ;;-- (defun get-keyword-val (keyword list) (second (member keyword list))) ;; this can't go in psqllex #+:null (defun lexdb-enabled-p nil (when *lexdb-params* (cond ((member :psql *features*) t) (t (format t "~%WARNING: ignoring *lexdb-params* (distribution is not :psql-enabled)") nil)))) ;; this can't go in psqllex #+:null (defun mwe-lexicon-enabled-p nil (and (lexdb-enabled-p) *postgres-mwe-enable*)) ;;; ;;; format conversion ;;; (defun str-2-symb (str) (unless (stringp str) (error "string exected")) (intern (string-upcase str) :lkb)) (defun str-2-keyword (str) (unless (stringp str) (error "string exected")) (intern str :keyword)) ;(defun str-2-keyword (str) ; (unless (stringp str) ; (error "string exected")) ; (intern (string-upcase str) :keyword)) ;; currently 'str-2-lisp-object' ... (defun str-2-list (str) (with-package (:lkb) (let ((item (read-from-string str))) (cond ((listp item) item) (t (error "list expected")))))) ;; use (parse-integer X :junk-allowed t) for integers (defun str-2-num (str &optional default) (with-package (:lkb) (let ((item (read-from-string str))) (cond ((numberp item) item) ((eq default t) (error "number expected")) (t default))))) (defun str-2-numstr (str &optional default) (let ((num (str-2-num str))) (cond ((numberp num) (format nil "~a" num)) ((eq default t) (error "number expected")) (t default)))) (defun explode-to-chars (string) (loop for i from 0 to (1- (length string)) collect (aref string i))) (defun escape-char (esc-char string) (implode-from-chars (loop for i from 0 to (1- (length string)) for c = (aref string i) if (or (eq c esc-char) (eq c #\\)) collect #\\ collect c))) #+:psql (defun escape-sql-copy-string (esc-char string) (implode-from-chars (loop for i from 0 to (1- (length string)) for c = (aref string i) if (eq c #\Backspace) collect #\\ and collect #\b else if (eq c #\Page) collect #\\ and collect #\f else if (eq c #\Return) collect #\\ and collect #\r else if (eq c #\Newline) collect #\\ and collect #\n else if (eq c #\Tab) collect #\\ and collect #\t else if (eq c #\vt) collect #\\ and collect #\v else if (eq c #\\) collect #\\ and collect #\\ else if (eq c esc-char) collect #\\ and collect c else collect c))) (defun replace-char (string &key from to) (unless (and (characterp from) (characterp to)) (error ":from and :to must be characters")) (implode-from-chars (loop for i from 0 to (1- (length string)) for c = (aref string i) if (eq c from) collect to else collect c))) (defun implode-from-chars (char-list) (loop with res = (make-string (length char-list)) for c in char-list for i upfrom 0 do (setf (schar res i) c) finally (return res))) ;; map string list to CVS copy line (defun str-list-2-line (str-list &key (sep-c #\Tab) (null-str "\\N")) (unless (listp str-list) (error "list argument expected")) (let ((sep (string sep-c))) (cond ;; return empty line if empty string list ((null str-list) "") ;; build line ;; sep-c separates values ;; values are null-str or escape-sql-copy-string'd vals (t (apply 'concatenate (cons 'string (cons (escape-sql-copy-string sep-c (pop str-list)) (mapcan #'(lambda (x) (list sep (if x (escape-sql-copy-string sep-c x) null-str) )) str-list)))))))) (defun str-list-2-str (str-list &key (sep-c #\Space) (null-str ":null:") (esc t)) (unless (listp str-list) (error "list expected")) (let ((sep (string sep-c))) (cond ((null str-list) "") (t (apply 'concatenate (cons 'string (cons (if esc (escape-char sep-c (pop str-list)) (pop str-list)) (mapcan #'(lambda (x) (list sep (if x (if esc (escape-char sep-c x) x) null-str) )) str-list)))))))) (defun str-list-2-str-by-str (str-list &optional (separator " ")) (unless (listp str-list) (error "list expected")) (cond ((null str-list) "") (t (apply 'concatenate (cons 'string (cons (pop str-list) (mapcan #'(lambda (x) (list separator x)) str-list))))))) (defun symb-2-str (symb) (unless (symbolp symb) (error "symbol expected")) (cond ((null symb) "") (t (string-downcase (string symb))))) (defun num-2-str (num) (if (null num) (return-from num-2-str)) (unless (numberp num) (error "number expected")) (format nil "~a" num)) (defun char-2-symb (c) (unless (characterp c) (error "character expected")) (str-2-symb (string c))) (defun char-2-num (c) (unless (characterp c) (error "character expected")) (str-2-num (string c))) (defun 2-symb (x) (cond ((symbolp x) x) ((stringp x) (str-2-symb x)) (t (error "unhandled type")))) (defun 2-str (x) (cond ((stringp x) x) ((symbolp x) (symb-2-str x)) ((numberp x) (num-2-str x)) ((pathnamep x) (namestring x)) (t (error "unhandled type")))) (defun null-or-2-str (x) (cond ((null x) nil) ((stringp x) x) ((symbolp x) (symb-2-str x)) ((numberp x) (num-2-str x)) (t (error "unhandled type")))) ;;; ;;; misc ;;; (defun make-nice-temp-file-pathname (filename) (make-pathname :name filename :host (pathname-host (lkb-tmp-dir)) :device (pathname-device (lkb-tmp-dir)) :directory (pathname-directory (lkb-tmp-dir)))) (defun get-grammar-version nil (let ((grammar-version (or (and (find-package :lkb) (find-symbol "*GRAMMAR-VERSION*" :lkb)) (find-symbol "*GRAMMAR-VERSION*" :common-lisp-user)))) (if (and grammar-version (boundp grammar-version)) (remove-if-not #'alphanumericp (symbol-value grammar-version)) "UNKNOWN_VERSION"))) (defun lex (str) (car (member str (extra-lexicons *lexicon*) :key #'name :test #'string=))) #+:null (defun recomp (x) (compile-file x) (load x)) ;;; ;;; initial code to support a notion of orthographic variants; see comments on ;;; *duplicate-lex-ids* in `mrs/generate.lisp'. quite possibly, this would be ;;; easier to do in the DB natively, i.e. comparing relevant field values, but ;;; this version at least i can run without a DB (or with). (8-aug-04; oe) ;;; ;;; ;;; a global repository of orthographic variants, relating sets of lexical ids ;;; that are equivalent modulo their STEM and ONSET values. ;;; (defparameter *orthographic-variants* nil) ;;; ;;; mostly for efficiency reasons: extract a hash key per lexical entry, such ;;; that we need only compare entries with identical keys. an invalid path for ;;; a non-Matrix grammar should result in loss of efficiency but not desaster. ;;; (defparameter *key-path* '(SYNSEM LOCAL CAT HEAD KEYS KEY)) ;;; ;;; _fix_me_ ;;; give some thought to memory management, in the sense that we will end up ;;; with the full (expanded) lexicon in core, unless the caching in the lexicon ;;; itself did something clever to limit cache growth. the latter, i guess, ;;; might be a good thing to aim for, if not available already. (8-aug-04; oe) ;;; (defun find-orthographic-variants (&key file) (let ((stream (if (stringp file) (open file :direction :output :if-exists :supersede) t)) (map (make-hash-table :test #'equal)) classes) ;; ;; first, populate .map. with pairs of lexical entries and restricted AVMs, ;; indexed by the type at *key-path* to reduce the number of comparisons. ;; (loop with *packing-restrictor* = (cons 'onset (last *orth-path*)) for id in (collect-psort-ids *lexicon*) for entry = (get-lex-entry-from-id id) for dag = (tdfs-indef (lex-entry-full-fs entry)) for key = (let ((dag (existing-dag-at-end-of dag *key-path*))) (when dag (dag-type dag))) for copy = (copy-dag-partially dag) do (push (cons entry copy) (gethash key map))) ;; ;; next, compare all pairs (of pairs) in each .map. bucket against each ;; other, though making sure to not do all pairs twice; records classes of ;; two or more lexical entries that compare equivalent (in their restricted ;; AVMs, i.e. sans orthography). ;; (loop for bucket being each hash-value in map do (loop with matches for entry = (pop bucket) for dag1 = (rest entry) for class = (list (first entry)) while entry do (loop for (key . dag2) in bucket when (dag-equal-p dag1 dag2) unless (member key matches :test #'eq) do (push key matches) (push key class)) (when (rest class) (push class classes)))) (setf *orthographic-variants* (make-hash-table)) ;; ;; finally, do two things: set the global variable *orthographic-variants* ;; and output a textual representation of orthographic variants for later ;; re-use. within each class sort lexicographically (for lack of a better ;; notion of precedence), but respect an original list of suppressed items ;; in *duplicate-lex-ids*, in case the grammar still supplies one. ;; (loop for class in classes do (loop with entries = (sort class #'string-lessp :key #'lex-entry-id) for entry in entries for id = (lex-entry-id entry) when (member id *duplicate-lex-ids* :test #'eq) collect id into passive else collect id into active finally (format stream "~@[~{~(~a~)~^ ~}~]~@[ : ~{~(~a~)~^ ~}~]~%" active passive) (loop for entry in entries for id = (lex-entry-id entry) for variants = (loop for foo in entries for bar = (lex-entry-id foo) unless (eq bar id) collect bar) do (setf (gethash id *orthographic-variants*) variants)))) (when (and (stringp file) (open-stream-p stream) (close stream))) classes)) (defun orthographic-variants-p (le1 le2) (when (hash-table-p *orthographic-variants*) (let ((le1 (intern le1 :lkb)) (le2 (intern le2 :lkb))) (or (eq le1 le2) (member le1 (gethash le2 *orthographic-variants*) :test #'eq) (member le2 (gethash le1 *orthographic-variants*) :test #'eq)))))