Malayalam has case, but not agreement. I chose to work on Case, lexical rules for number and case, lexical rules for verb tense, and started to work on adjectives. CASE ======================================================================= Malayalam uses case to indicate the roles of the nouns in the sentence. There are 7 cases: Nominative, Accusative, Dative, Sociative, Locative, Instrumental and Genative. My test cases primarily use Nominative, Accusative and Dative cases. The case is indicated by adding a suffix to the noun. Nominative case adds nothing. NOM: - ACC: -(y)e DAT: -kke / -(n)x The case system for pronouns is the same as for nouns. There are the same number of cases. The generation of the lexical entries for the different cases can be quite irregular for the pronouns. These bound suffixes are used for several different markers, and the order in which they apply to nouns is: stem (+ derivational suffix) (+plural) + case. None of my test cases use derivational forms of the nouns, so my lexical rules only handle number and case. The stem itself may have an associated gender. I modified malayalam.tdl to subdivide noun-lex (derived from basic-noun-lex) into common-noun-lex and pronoun-lex. The lexical entries for pronoun-lex are fully specified since the rules to generate them are irregular in some cases. The lexical rules to generate case are then written to modify only common-nouns. The lexical type for common-noun specifies that they are third person, so the entries for the common nouns are written to specify their GEND value, and I added an INFLECTION - to facilitate tracking of inflection by the rules. The lexical rule for number is applied first, since that is the first suffix applied to the stem. (described below) The noun-case-lex-rule written to add the case information takes as input the output from that rule. There are sub rules to add the various suffixes. The nominative rule is a constant-lex-rule since it adds no suffix. The accusative and dative rules are inflecting. The accusative rule actually has 2 possible suffixes: -ye is added if the word ends in a vowel, otherwise -e is added. For the dative case, all of my test cases using dative form use pronouns, so I have a simplified form of the dative inflection rule which adds the most common form: kkx to the stem. NUMBER ======================================================================= The lexical rule to inflect nouns for number are written to apply only to common nouns, since the pronouns are already fully specified. There are 2 possible number values: singular and plural. The suffix which seems to be most common for plural nouns is -kala. The plural suffix is added to the noun stem first, so the lexical rule is written to accept only common nouns with INFLECTION -. It's a lexeme-to-lexeme rule since the output will be subjected to a case inflection. Singular nouns have no changes, so the singular noun rule only adds the information in the INDEX.PNG feature that NUM is singular. Plural nouns are inflected to add the -kala suffix, and the INDEX.PNG feature is set to NUM non-singular. TENSE ======================================================================= Malayalam verbs are inflected for tense by adding a suffix to the verb stem. The suffixes are: Past -i / -u Present -unnu Future - um My test cases only include examples of transitive verbs, so I wrote the lexical rule as trans-verb-tense-lex-rule := lexeme-to-word-rule & [DTR transitive-verb-lex]. Then I added 3 inflection rules: past-trans-verb-rule, which adds a -u suffix when the word ends in a double consonant, otherwise adds an -i pres-trans-verb-rule, which adds a -unnu suffix futr-trans-verb-rule, which adds a -um suffix ADJECTIVES ======================================================================= Adjectives may be inflected to match the number and gender of the noun they modify. The adjective 'good' in my test cases is marked by a suffixes in the following manner: + HUMAN SINGULAR MASC -van FEM - vala PLURAL - var - HUMAN SINGULAR - tx PLURAL - va I copied the English adjective example, but do not have it working yet. In Malayalam, the adjective follows the noun. GENERATION ======================================================================= Generation produced many, many sentences! I get sentences with both word orders for subject and object, which is expected. I also get all of the verb tenses. The one area which doesn't look right is the -oo suffix used in matrix questions. It should be added as a suffix to the last word in the sentence, but instead it's being added as a separate word. I'll need to fix the rule for that. QUESTIONS ======================================================================= There are some words which are already inflected, since they have irregular forms. How do I write rules with multiple inflections, which case take these as input partway through the inflection? Or do I just write a fully inflected form for them? The example I have is snake, for which the plural form seems to be irregular. After the plural form however, I could let it inflect for case following the normal rules. There are some sentences where both the subject and object are in nominative case. I'm not sure what the rules are which determine which case should be used. The lexical rules are currently written so that verbs take a nominative and accusative noun, but this will have to be revised to make it much more open. How can I control overgeneration in this case? What are some strategies?