French stemming algorithm

Links to resources

• Romance language stemmers
• Javascript demo
• The stemmer in Snowball
• Sample French vocabulary
• Its stemmed equivalent
• French stop word list (UTF-8 encoding)

Here is a sample of French vocabulary, with the stemmed forms that will be generated by this algorithm:

The stemming algorithm

Letters in French include the following accented forms,

â   à   ç   ë   é   ê   è   ï   î   ô   û   ù

The following letters are vowels:

a   e   i   o   u   y   â   à   ë   é   ê   è   ï   î   ô   û   ù

The first step removes elisions. If the word starts with one of c d j l m n s t or qu, followed by an apostrophe (') then remove this unless it comprises the whole word.

Assume the word is in lower case. Then, taking the letters in turn from the beginning to end of the word, put u or i into upper case when it is both preceded and followed by a vowel; put y into upper case when it is either preceded or followed by a vowel; and put u into upper case when it follows q. For example,

jouer		→		joUer
ennuie		→		ennuIe
yeux		→		Yeux
quand		→		qUand
croyiez		→		croYiez

In the last example, y becomes Y because it is between two vowels, but i does not become I because it is between Y and e, and Y is not defined as a vowel above.

(The upper case forms are not then classed as vowels — see note on vowel marking.)

Replace ë and ï with He and Hi. The H marks the vowel as having originally had a diaeresis, while the vowel itself, lacking an accent, is able to match suffixes beginning in e or i.

If the word begins with two vowels, RV is the region after the third letter, otherwise the region after the first vowel not at the beginning of the word, or the end of the word if these positions cannot be found. (Exceptionally, par, col or tap, at the beginning of a word is also taken to define RV as the region to their right.)

For example,

    a i m e r     a d o r e r     v o l e r    t a p i s
         |...|         |.....|       |.....|        |...|

R1 is the region after the first non-vowel following a vowel, or the end of the word if there is no such non-vowel.

R2 is the region after the first non-vowel following a vowel in R1, or the end of the word if there is no such non-vowel. (See note on R1 and R2.)

For example:

    f a m e u s e m e n t
         |......R1.......|
               |...R2....|

Note that R1 can contain RV (adorer), and RV can contain R1 (voler).

Below, ‘delete if in R2’ means that a found suffix should be removed if it lies entirely in R2, but not if it overlaps R2 and the rest of the word. ‘delete if in R1 and preceded by X’ means that X itself does not have to come in R1, while ‘delete if preceded by X in R1’ means that X, like the suffix, must be entirely in R1.

Start with step 1

Step 1: Standard suffix removal

Search for the longest among the following suffixes, and perform the action indicated.

ance   iqUe   isme   able   iste   eux   ances   iqUes   ismes   ables   istes

delete if in R2

atrice   ateur   ation   atrices   ateurs   ations

delete if in R2

if preceded by ic, delete if in R2, else replace by iqU

logie   logies

replace with log if in R2

usion   ution   usions   utions

replace with u if in R2

ence   ences

replace with ent if in R2

ement   ements

delete if in RV

if preceded by iv, delete if in R2 (and if further preceded by at, delete if in R2), otherwise,

if preceded by eus, delete if in R2, else replace by eux if in R1, otherwise,

if preceded by abl or iqU, delete if in R2, otherwise,

if preceded by ièr or Ièr, replace by i if in RV

ité   ités

delete if in R2

if preceded by abil, delete if in R2, else replace by abl, otherwise,

if preceded by ic, delete if in R2, else replace by iqU, otherwise,

if preceded by iv, delete if in R2

if   ive   ifs   ives

delete if in R2

if preceded by at, delete if in R2 (and if further preceded by ic, delete if in R2, else replace by iqU)

eaux

replace with eau

aux

replace with al if in R1

euse   euses

delete if in R2, else replace by eux if in R1

issement   issements

delete if in R1 and preceded by a non-vowel

amment

replace with ant if in RV

emment

replace with ent if in RV

ment   ments

delete if preceded by a vowel in RV

In steps 2a and 2b all tests are confined to the RV region.

Do step 2a if either no ending was removed by step 1, or if one of endings amment, emment, ment, ments was found.

Step 2a: Verb suffixes beginning i

Search for the longest among the following suffixes and if found, delete if the preceding character is neither a vowel nor H.

îmes   ît   îtes   i   ie   ies   ir   ira   irai   iraIent   irais   irait   iras   irent   irez   iriez   irions   irons   iront   is   issaIent   issais   issait   issant   issante   issantes   issants   isse   issent   isses   issez   issiez   issions   issons   it

(Note that the preceding character itself must also be in RV.)

Do step 2b if step 2a was done, but failed to remove a suffix.

Step 2b: Other verb suffixes

Search for the longest among the following suffixes, and perform the action indicated.

ions

delete if in R2

é   ée   ées   és   èrent   er   era   erai   eraIent   erais   erait   eras   erez   eriez   erions   erons   eront   ez   iez

delete

âmes   ât   âtes   a   ai   aIent   ais   ait   ant   ante   antes   ants   as   asse   assent   asses   assiez   assions

delete

if preceded by e, delete

(Note that the e that may be deleted in this last step must also be in RV.)

If the last step to be obeyed — either step 1, 2a or 2b — altered the word, do step 3

Step 3

Replace final Y with i or final ç with c

Alternatively, if the last step to be obeyed did not alter the word, do step 4

Step 4: Residual suffix

If the word ends s, not preceded by a, i (unless itself preceded by H), o, u, è or s, delete it.

In the rest of step 4, all tests are confined to the RV region.

Search for the longest among the following suffixes, and perform the action indicated.

ion

delete if in R2 and preceded by s or t

ier   ière   Ier   Ière

replace with i

e

delete

(So note that ion is removed only when it is in R2 — as well as being in RV — and preceded by s or t which must be in RV.)

Always do steps 5 and 6.

Step 5: Undouble

If the word ends enn, onn, ett, ell or eill, delete the last letter

Step 6: Un-accent

If the words ends é or è followed by at least one non-vowel, remove the accent from the e.

And finally:

Turn any remaining I, U and Y letters in the word back into lower case.

Turn He and Hi back into ë and ï, and remove any remaining H.

History of functional changes to the algorithm

• Snowball 2.0.0: Suffixes that begin with a diaereses are now removed (done by replacing ë and ï with He and Hi, during stemming then undoing afterwards).
• Snowball 2.3.0: Added a new first step which removes elisions.

The same algorithm in Snowball

routines (
           elisions
           prelude postlude mark_regions
           RV R1 R2
           standard_suffix
           i_verb_suffix
           verb_suffix
           residual_suffix
           un_double
           un_accent
)

externals ( stem )

integers ( pV p1 p2 )

groupings ( elision_char v keep_with_s )

stringescapes {}

/* special characters */

stringdef a^   '{U+00E2}'  // a-circumflex
stringdef a`   '{U+00E0}'  // a-grave
stringdef cc   '{U+00E7}'  // c-cedilla

stringdef e"   '{U+00EB}'  // e-diaeresis (rare)
stringdef e'   '{U+00E9}'  // e-acute
stringdef e^   '{U+00EA}'  // e-circumflex
stringdef e`   '{U+00E8}'  // e-grave
stringdef i"   '{U+00EF}'  // i-diaeresis
stringdef i^   '{U+00EE}'  // i-circumflex
stringdef o^   '{U+00F4}'  // o-circumflex
stringdef u^   '{U+00FB}'  // u-circumflex
stringdef u`   '{U+00F9}'  // u-grave

define v 'aeiouy{a^}{a`}{e"}{e'}{e^}{e`}{i"}{i^}{o^}{u^}{u`}'

// Single character elisions
define elision_char 'cdjlmnst'

define elisions as
(
    [ (elision_char or 'qu') '{'}' ] not atlimit delete
)

define prelude as repeat goto (

    (  v [ ('u' ] v <- 'U') or
           ('i' ] v <- 'I') or
           ('y' ] <- 'Y')
    )
    or
    (  [ '{e"}' ] <- 'He' )
    or
    (  [ '{i"}' ] <- 'Hi' )
    or
    (  ['y'] v <- 'Y' )
    or
    (  'q' ['u'] <- 'U' )
)

define mark_regions as (

    $pV = limit
    $p1 = limit
    $p2 = limit  // defaults

    do (
        ( v v next )
        or
        among ( // this exception list begun Nov 2006
            'par'  // paris, parie, pari
            'col'  // colis
            'tap'  // tapis
            // extensions possible here
        )
        or
        ( next gopast v )
        setmark pV
    )
    do (
        gopast v gopast non-v setmark p1
        gopast v gopast non-v setmark p2
    )
)

define postlude as repeat (

    [substring] among(
        'I' (<- 'i')
        'U' (<- 'u')
        'Y' (<- 'y')
        'He' (<- '{e"}')
        'Hi' (<- '{i"}')
        'H' (delete)
        ''  (next)
    )
)

backwardmode (

    define RV as $pV <= cursor
    define R1 as $p1 <= cursor
    define R2 as $p2 <= cursor

    define standard_suffix as (
        [substring] among(

            'ance' 'iqUe' 'isme' 'able' 'iste' 'eux'
            'ances' 'iqUes' 'ismes' 'ables' 'istes'
               ( R2 delete )
            'atrice' 'ateur' 'ation'
            'atrices' 'ateurs' 'ations'
               ( R2 delete
                 try ( ['ic'] (R2 delete) or <-'iqU' )
               )
            'logie'
            'logies'
               ( R2 <- 'log' )
            'usion' 'ution'
            'usions' 'utions'
               ( R2 <- 'u' )
            'ence'
            'ences'
               ( R2 <- 'ent' )
            'ement'
            'ements'
            (
                RV delete
                try (
                    [substring] among(
                        'iv'   (R2 delete ['at'] R2 delete)
                        'eus'  ((R2 delete) or (R1<-'eux'))
                        'abl' 'iqU'
                               (R2 delete)
                        'i{e`}r' 'I{e`}r'      //)
                               (RV <-'i')      //)--new 2 Sept 02
                    )
                )
            )
            'it{e'}'
            'it{e'}s'
            (
                R2 delete
                try (
                    [substring] among(
                        'abil' ((R2 delete) or <-'abl')
                        'ic'   ((R2 delete) or <-'iqU')
                        'iv'   (R2 delete)
                    )
                )
            )
            'if' 'ive'
            'ifs' 'ives'
            (
                R2 delete
                try ( ['at'] R2 delete ['ic'] (R2 delete) or <-'iqU' )
            )
            'eaux' (<- 'eau')
            'aux'  (R1 <- 'al')
            'euse'
            'euses'((R2 delete) or (R1<-'eux'))

            'issement'
            'issements'(R1 non-v delete) // verbal

            // fail(...) below forces entry to verb_suffix. -ment typically
            // follows the p.p., e.g 'confus{e'}ment'.

            'amment'   (RV fail(<- 'ant'))
            'emment'   (RV fail(<- 'ent'))
            'ment'
            'ments'    (test(v RV) fail(delete))
                       // v is e,i,u,{e'},I or U
        )
    )

    define i_verb_suffix as setlimit tomark pV for (
        [substring] among (
            '{i^}mes' '{i^}t' '{i^}tes' 'i' 'ie' 'ies' 'ir' 'ira' 'irai'
            'iraIent' 'irais' 'irait' 'iras' 'irent' 'irez' 'iriez'
            'irions' 'irons' 'iront' 'is' 'issaIent' 'issais' 'issait'
            'issant' 'issante' 'issantes' 'issants' 'isse' 'issent' 'isses'
            'issez' 'issiez' 'issions' 'issons' 'it'
                (not 'H' non-v delete)
        )
    )

    define verb_suffix as setlimit tomark pV for (
        [substring] among (
            'ions'
                (R2 delete)

            '{e'}' '{e'}e' '{e'}es' '{e'}s' '{e`}rent' 'er' 'era' 'erai'
            'eraIent' 'erais' 'erait' 'eras' 'erez' 'eriez' 'erions'
            'erons' 'eront' 'ez' 'iez'

            // 'ons' //-best omitted

                (delete)

            '{a^}mes' '{a^}t' '{a^}tes' 'a' 'ai' 'aIent' 'ais' 'ait' 'ant'
            'ante' 'antes' 'ants' 'as' 'asse' 'assent' 'asses' 'assiez'
            'assions'
                (delete
                 try(['e'] delete)
                )
        )
    )

    define keep_with_s 'aiou{e`}s'

    define residual_suffix as (
        try(['s'] test ('Hi' or non-keep_with_s) delete)
        setlimit tomark pV for (
            [substring] among(
                'ion'           (R2 's' or 't' delete)
                'ier' 'i{e`}re'
                'Ier' 'I{e`}re' (<-'i')
                'e'             (delete)
            )
        )
    )

    define un_double as (
        test among('enn' 'onn' 'ett' 'ell' 'eill') [next] delete
    )

    define un_accent as (
        atleast 1 non-v
        [ '{e'}' or '{e`}' ] <-'e'
    )
)

define stem as (

    do elisions
    do prelude
    do mark_regions
    backwards (

        do (
            (
                 ( standard_suffix or
                   i_verb_suffix or
                   verb_suffix
                 )
                 and
                 try( [ ('Y'   ] <- 'i' ) or
                        ('{cc}'] <- 'c' )
                 )
            ) or
            residual_suffix
        )

        // try(['ent'] RV delete) // is best omitted

        do un_double
        do un_accent
    )
    do postlude
)