Text language detection

• US 7,035,801 B2
• Filed: 09/05/2001
• Issued: 04/25/2006
• Est. Priority Date: 09/06/2000
• Status: Active Grant

• Current Assignee: Telefonaktiebolaget LM Ericsson
• Original Assignee: Telefonaktiebolaget LM Ericsson
• Inventors: Jimenez-Feltström, Alberto
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US09/945,663
• Publication Number:
  US 20020095288A1
• Time in Patent Office: 1,693 Days
• Field of Search: 704/260, 704/277, 704/257, 704/9, 704/2, 704/278, 717/1, 382/229, 382/228, 395/600, 395/757
• US Class Current: 704/255
• CPC Class Codes: G06F 40/263 Language identification
• CPC Class Codes: G10L 13/08 Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination

Abstract

A method of determining the language of a text message received by a mobile telecommunications device indicates receiving an input text message at a mobile telecommunications device; analyzing the input text message using language information stored in the mobile telecommunications device; selecting, from a group of languages defined by the language information, a most likely language for the input text message; and outputting, from the mobile telecommunications device, speech signals corresponding to the input text message, in the selected language.

Claims

1. A method of determining the language of a text message that comprises a plurality of words, the method comprising:

receiving an input text message consisting of a plurality of characters;

dividing the plurality of characters into a series of character groups, each consisting of n characters;

for each character group, except the last in the message, determining respective probability values for selected languages, each probability value being indicative of the probability that the character group concerned is followed by the next character group in the message in the language concerned;

for each language, accumulating the probability values determined for each character group in the message; and

determining the language of the text message on the basis of the accumulated probability values.

2. A method as claimed in claim 1, wherein probability values are pre-stored in storage means and are retrieved therefrom.

3. A method as claimed in claim 1, wherein the probability values are determined by transition probability values of a discrete Markov model of each language.

4. A method as claimed in claim 3, wherein, for each language, the probability values are accumulated according to the equation:

${\widehat{\mathrm{\Lambda <br><br>}}}_m=p<br><br>\left({\mathrm{\Lambda <br><br>}}_m\right)<br><br>q_{u_1}^{\left(m\right)}<br><br>\underset{i=1}{\overset{N-1}{\prod <br><br>}}<br><br><br><br>a_{u_i,u_{i+1}}^{\left(m\right)}$

in which {circumflex over (Λ

)}_m=accumulated probability for language m,

p(Λ

_m)=a-priori probability for language m,

q_ui^(m)=probability that the text message begins with character group u_i for language m,

N=number of character groups in the text message, and

α

_ui,ui+1^(m)=probability that character group u_i+1 follows character group u_i in language m.

5. A method as claimed in claim 3, wherein for each language, the probability values are stored as probability values.

6. A method as claimed in claim 5, wherein, for each language, the probability values are accumulated according to the equation:

${\widehat{\mathrm{\Lambda <br><br>}}}_m=\log <br><br>\left(p<br><br>\left({\mathrm{\Lambda <br><br>}}_m\right)\right)+\log <br><br>\left(q_{u_1}^{\left(m\right)}\right)+\underset{i=1}{\overset{N-1}{\sum <br><br>}}<br><br><br><br>\log <br><br>\left(a_{u_i,u_{i+1}}^{\left(m\right)}\right)$

in which {circumflex over (Λ

)}_m=accumulated probability for language m,

log(p(Λ

_m))=logarithm of a-priori probability for language m,

log(q_ui^(m))=logarithm of probability that text message begins with group u_i for language m,

N=number of character groups in the text message,

log(α

_ui,ui+1^(m)=logarithm of probability that character group u_i+1 of the text message follows character group u_i in language m.

7. A method as claimed in claim 1, wherein the number, n, of characters in each group is 2.

8. A method as claimed in claim 1, wherein the number, n, of characters in each group is 3.

9. A method as claimed in claim 1, wherein the number n, of characters in each group, is 1.