WeightedFiniteStateTransducers
WeightedFiniteStateTransducers
  • 13
  • 70 180

วีดีโอ

Lecture 12 Introduction to 5 Basic Operations for WFST Weight Pushing
มุมมอง 2K9 ปีที่แล้ว
Download link: www.dropbox.com/s/k8oj4q72co8jg4g/Lecture_12 Introduction_to_5_Basic_Operations_for_WFST Weight_Pushing.pdf?dl=0
Lecture 11 Introduction to 5 Basic Operations for WFST Determinization
มุมมอง 2.6K9 ปีที่แล้ว
Download link: www.dropbox.com/s/r5ozvoe1yqch33b/Lecture_11 Introduction_to_5_Basic_Operations_for_WFST Determinization.pdf?dl=0
Lecture 10 Introduction to 5 Basic Operations for WFST Epsilon Removal
มุมมอง 2.3K9 ปีที่แล้ว
Download link: www.dropbox.com/s/yxz7j2xkhi9alph/Lecture_10 Introduction_to_5_Basic_Operations_for_WFST Epsilon_Removal.pdf?dl=0
Lecture 8 Introduction to Basic Operations for Finite Automaton
มุมมอง 3.1K9 ปีที่แล้ว
Download link: www.dropbox.com/s/gpwul8y1khjrxyd/Lecture_8 Introduction_to_Basic_Operations_for_Finite_Automaton.pdf?dl=0
Lecture 7 Introduction to Semirings V
มุมมอง 2.8K9 ปีที่แล้ว
Download link: www.dropbox.com/s/rdtlwva0epttzon/Lecture_7 Introduction_to_Semirings_V.pdf?dl=0
Lecture 6 Introduction to Semirings IV
มุมมอง 2.6K9 ปีที่แล้ว
Download link: www.dropbox.com/s/bnssx8soundqmo5/Lecture_6 Introduction_to_Semirings_IV.pdf?dl=0
Lecture 5 Introduction to Semirings III
มุมมอง 3.7K9 ปีที่แล้ว
Download link: www.dropbox.com/s/vfzlectvb8ftdys/Lecture_5 Introduction_to_Semirings_III.pdf?dl=0
Lecture 4 Introduction to Semirings II
มุมมอง 5K9 ปีที่แล้ว
Download link: www.dropbox.com/s/8opwx45uuwqjth5/Lecture_4 Introduction_to_Semirings_II.pdf?dl=0
Lecture 3 Introduction to Semirings I
มุมมอง 10K9 ปีที่แล้ว
Download link: www.dropbox.com/s/54bvdzxtnivo6vw/Lecture_3 Introduction_to_Semirings_I.pdf?dl=0
Lecture 2 Introduction to Finite State Transducers
มุมมอง 16K9 ปีที่แล้ว
Download link: www.dropbox.com/s/0774w4b7vw99gmr/Lecture_2 Introduction_to_Finite_State_Transducers.pdf?dl=0
Lecture 1 Introduction to Finite Automaton
มุมมอง 12K9 ปีที่แล้ว
Download link: www.dropbox.com/s/pvjodr3qdbmmvk5/Lecture_1 Introduction_to_Finite_Automaton.pdf?dl=0
Lecture 9 Introduction to 5 Basic Operations for WFST Composition
มุมมอง 5K9 ปีที่แล้ว
Download link: www.dropbox.com/s/mbkpbv2ndb1jc7o/Lecture_9 Introduction_to_5_Basic_Operations_for_WFST Composition.pdf?dl=0

ความคิดเห็น

  • @rossroessler5159
    @rossroessler5159 หลายเดือนก่อน

    Thank you so much, it's the first explanation of how to actually do composition that I've seen.

  • @reemgody5492
    @reemgody5492 6 หลายเดือนก่อน

    very nice explanation

  • @ananthakrishnank3208
    @ananthakrishnank3208 ปีที่แล้ว

    Many thanks! I am going through hbka.pdf (WFST paper); found difficult to grasp just from notations without an example. The way of example always helps.

  • @waisyousofi9139
    @waisyousofi9139 ปีที่แล้ว

    Thanks for the great content, can you leave the link to download Hori's book

  • @kimchi_taco
    @kimchi_taco ปีที่แล้ว

    It would be more great if it mentions Tropical is used for edit distance computation, and Log is used for probability and entropy computation.

  • @hardikkamboj3528
    @hardikkamboj3528 2 ปีที่แล้ว

    2:42, the first string would not be accepted. This machine only checks the pattern "10" in the last 2 characters of the input.

  • @buyuancui4468
    @buyuancui4468 2 ปีที่แล้ว

    Thank you!

  • @erayser
    @erayser 3 ปีที่แล้ว

    Very well explained! Thank you!!

  • @siddiquifahad85
    @siddiquifahad85 3 ปีที่แล้ว

    Thank you for the lecturers. They helped me greatly !

  • @julianfortune8974
    @julianfortune8974 3 ปีที่แล้ว

    Thank you!

  • @CyznK
    @CyznK 3 ปีที่แล้ว

    maybe there is a typo in Example 1 1 x 1 -> 1 + 1>

  • @buoyrina9669
    @buoyrina9669 3 ปีที่แล้ว

    Thanks for such a good content

  • @aanchan
    @aanchan 3 ปีที่แล้ว

    There is a bug on the fourth iteration slide, the table in the first column Input/Output/Weight of A should say 2_A and not 3_A

    • @stanstarner3249
      @stanstarner3249 10 หลายเดือนก่อน

      3_A is correct.

    • @stanstarner3249
      @stanstarner3249 10 หลายเดือนก่อน

      He's starting from 2_A and the edge is from 2_A to 3_A.

  • @Sung-LinYeh
    @Sung-LinYeh 3 ปีที่แล้ว

    Nice lectures! Thanks

  • @shujiehu1210
    @shujiehu1210 4 ปีที่แล้ว

    thanks a lot

  • @cyaaronk7328
    @cyaaronk7328 4 ปีที่แล้ว

    Wonderful lectures! Thank you very much!

  • @cyaaronk7328
    @cyaaronk7328 4 ปีที่แล้ว

    Amazing

  • @cyaaronk7328
    @cyaaronk7328 4 ปีที่แล้ว

    Excellent videos!

  • @amirworms1173
    @amirworms1173 4 ปีที่แล้ว

    Thank you very much for this lecture, very clear!

  • @lianhongw
    @lianhongw 4 ปีที่แล้ว

    Thank you Sir, very nice lectures.

  • @akshaydevkarama3277
    @akshaydevkarama3277 4 ปีที่แล้ว

    Really nice,your entire series helped me!

  • @ShahrukHossain
    @ShahrukHossain 4 ปีที่แล้ว

    Thank you for the lectures ! Throughly enjoyed and learned from them all !

  • @rickiest_rick7704
    @rickiest_rick7704 5 ปีที่แล้ว

    In the distributivity check at 7:00, why do you check the (z.x)+(z.y) = z.(x+y) , isn't that same as the first check?

  • @saminishtiaq
    @saminishtiaq 5 ปีที่แล้ว

    Typo in the slide: "Stochastic WFST: ...the sum of all transition *weights* leaving...". What you *say* is correct.

  • @jassay6435
    @jassay6435 5 ปีที่แล้ว

    I'm lost finally....

    • @saminishtiaq
      @saminishtiaq 5 ปีที่แล้ว

      The author is defining a new semiring, the Log Semiring. To do this, they have to give the domain (Reals, in this case), and the add and mult operators (with their units) over this domain. The mult operator is the normal + on their indices. The add operator, as the author says, is a bit funky. But they can't just say "here are the operators"; they have to prove that the operators have the properties (associativity, communtativity, identity; distribution of mult over add) that they need to define a semiring. The last few slides do the proof.

  • @kindianmguy
    @kindianmguy 5 ปีที่แล้ว

    nice explanation of FSA thanks

  • @jianpenghou8821
    @jianpenghou8821 5 ปีที่แล้ว

    Nice work! Thank you!

  • @jianpenghou8821
    @jianpenghou8821 5 ปีที่แล้ว

    What does the red check mark mean?

  • @thekishanpatel
    @thekishanpatel 5 ปีที่แล้ว

    Is the difference between a ring and a semiring that in the latter, the addition operator need not have the inverse property?

    • @thekishanpatel
      @thekishanpatel 5 ปีที่แล้ว

      NVM I just realized you went over this in the video...I had skipped to the end of the video and had missed that brief discussion

  • @fangjunkuang5061
    @fangjunkuang5061 5 ปีที่แล้ว

    Do you have many colleagues from India?

  • @fangjunkuang5061
    @fangjunkuang5061 5 ปีที่แล้ว

    Do you have many colleagues from India?

  • @fangjunkuang5061
    @fangjunkuang5061 5 ปีที่แล้ว

    epsilon is not defined 1:43 . You have used uppercase A to denote the set of input labels in slide 1 but 1:43 uses Sigma.

  • @fangjunkuang5061
    @fangjunkuang5061 5 ปีที่แล้ว

    The set of output labels Delta at 8:24 is not defined in your slides 4:06

  • @fangjunkuang5061
    @fangjunkuang5061 5 ปีที่แล้ว

    There's an error at 1:52 。The first string is invalid as it ends at state x1

  • @amr-keleg
    @amr-keleg 5 ปีที่แล้ว

    Why can the min for example be replaced with the addition operator? Could you please elaborate more?

    • @saminishtiaq
      @saminishtiaq 5 ปีที่แล้ว

      It can't be replaced. Suppose s = \bigoplus_{\pi \in P(S,T)} w(\pi). We still need to calculate min(s), or max(s), or union(s) to score the path.

  • @mohomedarshad7252
    @mohomedarshad7252 6 ปีที่แล้ว

    Hi, your videos are very concise and helpful. Can you also post videos related to tree transducers ?

  • @ulteriormotif
    @ulteriormotif 6 ปีที่แล้ว

    Very good video. Wouldn't a deterministic finite state transducer also need to not have any epsilon transitions as their presence would require a choice to be made to either follow a labelled transition and consume input or to follow the epsilon and consume nothing?

  • @JiaxingQi
    @JiaxingQi 6 ปีที่แล้ว

    I have a question. Why probability does not satisfy multiplication commutativity?

    • @Jawbreaker9202
      @Jawbreaker9202 5 ปีที่แล้ว

      I think that the semi ring definition states that it may or may not obey those properties, but that its not required to

    • @stanisawantonowicz8775
      @stanisawantonowicz8775 5 ปีที่แล้ว

      Because probability is defined on a unit interval (from 0 to one), so for example the multiplicative inverse of 1/2 would be 2, but 2 is not in a unit interval. So not for every element of a unit interval there exists a multiplicative inverse (only for 0 and 1 actually). Same goes for additive inverse, this time all inverses would be negative.

  • @linpershey
    @linpershey 6 ปีที่แล้ว

    a good entry video as the first step to understand FST (for people that understand states, transitions)

  • @musiclover19891
    @musiclover19891 6 ปีที่แล้ว

    These are very well explained, thank you!

  • @shaunliu5628
    @shaunliu5628 6 ปีที่แล้ว

    Good lectures!! THX!!!

  • @jaymefosa2003
    @jaymefosa2003 6 ปีที่แล้ว

    nice explanation style

  • @iktunutki
    @iktunutki 6 ปีที่แล้ว

    The worked example would be much easier to follow if states had alphabetical labels instead of numeric ones :( Everything is 0,1, 2

  • @iktunutki
    @iktunutki 6 ปีที่แล้ว

    The semiring videos were very clear and interesting and came with the idea that "they would be used later", but here I feel that the motivation for doing a composition (or other operations) hasn't been explained: what could we use it for?

  • @林建民-n8g
    @林建民-n8g 6 ปีที่แล้ว

    good job

  • @haiweiwu2814
    @haiweiwu2814 7 ปีที่แล้ว

    Wonderful lectures! THANK YOU VERY MUCH!

  • @xuedahu8504
    @xuedahu8504 7 ปีที่แล้ว

    感觉有点印度口音,但还是能听懂,哈哈,谢谢老师~

    • @xingcebao5379
      @xingcebao5379 4 ปีที่แล้ว

      这是正宗新加坡英语。。。

  • @张尚宇-i3u
    @张尚宇-i3u 7 ปีที่แล้ว

    Clear explanation

  • @duniandong2850
    @duniandong2850 7 ปีที่แล้ว

    a good help

  • @yujiaxiao428
    @yujiaxiao428 7 ปีที่แล้ว

    Thanks for your introduction! I have a question, Why the V[1] should be calculated as the sum of 1.5 and the special 0? There is only one path from state 1 to final state, why should we do the sum operation?