Loading ...
Sorry, an error occurred while loading the content.

Archeological Decipherment

Expand Messages
  • Michael Matsko
    Does anyone know of a more recent study of archeological decipherment than E. J. W. Barber s 1974 Archeological Decipherment? I am aware of Kevin Knight s
    Message 1 of 12 , Mar 29, 2014
    • 0 Attachment
      Does anyone know of a more recent study of archeological decipherment than E. J. W. Barber's 1974 Archeological Decipherment? I am aware of Kevin Knight's work (at USC's ISI) do not know of much else.

      Mike Matsko
      Arlington, VA
    • Peter T. Daniels
      What exactly are you looking for? She doesn t talk about any actual decipherments, but presents some mathematical algorithms. The standard work has been
      Message 2 of 12 , Mar 30, 2014
      • 0 Attachment
        What exactly are you looking for? She doesn't talk about any actual decipherments, but presents some mathematical algorithms. The standard work has been Maurice Pope's volume from Thames & Hudson (it had different titles in the UK and US editions, and for the second edition too), which is reliable on the whole except that in the 2nd edition a reference is added in a footnote to my article on Hincks's decipherment of Akkadian but clearly he didn't even look at it, since the text continues to ascribe the achievement to Rawlinson. (For the demonstration that Rawlinson contributed almost nothing to the decipherment of cuneiform, see my article in the on-line Encyclopedia Iranica, which is at iranica.org.)

        The journalist Andrew Robinson's Lost Languages is a coffee table book mostly about undeciphered scripts but includes some information on decipherments. (It's considerably better than his earlier History of Writing.) He also wrote little biographies of Michael Ventris and of Thomas Young.

        You'll find lots of references to the primary publications of decipherments in my chapter in The World's Writing Systems (1996), and overviews in the Oxford Encyclopedia of the Archeology of the Near East and in Civilizations of the Ancient Near East (Sasson et al.).. 
        --
        Peter T. Daniels grammatim@...


        From: Michael Matsko <msm@...>
        To: ANE-2@yahoogroups.com
        Sent: Saturday, March 29, 2014 10:37 PM
        Subject: [ANE-2] Archeological Decipherment

         
        Does anyone know of a more recent study of archeological decipherment than E. J. W. Barber's 1974 Archeological Decipherment? I am aware of Kevin Knight's work (at USC's ISI) do not know of much else.

        Mike Matsko
        Arlington, VA
      • Michael Matsko
        Peter, Thank you for your detailed and comprehensive response. To your question. I am looking more for references to algorithmic aids to decipherment than
        Message 3 of 12 , Mar 30, 2014
        • 0 Attachment
          Peter,

          Thank you for your detailed and comprehensive response.

          To your question.  I am looking more for references to algorithmic aids to decipherment than stories of decipherment.  An incredible amount of work has been done in computer science with respect to natural language processing in the last 40 years and I am interested in finding out if anyone in the ANE community looking at ancient languages is using any of these techniques.

          Mike Matsko

          On Mar 30, 2014, at 9:19 AM, "Peter T. Daniels" <grammatim@...> wrote:

           

          What exactly are you looking for? She doesn't talk about any actual decipherments, but presents some mathematical algorithms. The standard work has been Maurice Pope's volume from Thames & Hudson (it had different titles in the UK and US editions, and for the second edition too), which is reliable on the whole except that in the 2nd edition a reference is added in a footnote to my article on Hincks's decipherment of Akkadian but clearly he didn't even look at it, since the text continues to ascribe the achievement to Rawlinson. (For the demonstration that Rawlinson contributed almost nothing to the decipherment of cuneiform, see my article in the on-line Encyclopedia Iranica, which is at iranica.org.)

          The journalist Andrew Robinson's Lost Languages is a coffee table book mostly about undeciphered scripts but includes some information on decipherments. (It's considerably better than his earlier History of Writing.) He also wrote little biographies of Michael Ventris and of Thomas Young.

          You'll find lots of references to the primary publications of decipherments in my chapter in The World's Writing Systems (1996), and overviews in the Oxford Encyclopedia of the Archeology of the Near East and in Civilizations of the Ancient Near East (Sasson et al.).. 
          --
          Peter T. Daniels grammatim@...


          From: Michael Matsko <msm@...>
          To: ANE-2@yahoogroups.com
          Sent: Saturday, March 29, 2014 10:37 PM
          Subject: [ANE-2] Archeological Decipherment

           
          Does anyone know of a more recent study of archeological decipherment than E. J. W. Barber's 1974 Archeological Decipherment? I am aware of Kevin Knight's work (at USC's ISI) do not know of much else.

          Mike Matsko
          Arlington, VA

        • Peter T. Daniels
          Can you point to a sizable corpus of undeciphered texts where a computational algorithm might be useful? IIRC Barber offered a mathematicization of what Kober
          Message 4 of 12 , Mar 30, 2014
          • 0 Attachment
            Can you point to a sizable corpus of undeciphered texts where a computational algorithm might be useful? IIRC Barber offered a mathematicization of what Kober and Ventris had done with Linear B. 
            --
            Peter T. Daniels grammatim@...
            Jersey City


            From: Michael Matsko <msm@...>
            To: ANE-2@yahoogroups.com
            Sent: Sunday, March 30, 2014 11:16 AM
            Subject: Re: [ANE-2] Archeological Decipherment

             
            Peter,

            Thank you for your detailed and comprehensive response.

            To your question.  I am looking more for references to algorithmic aids to decipherment than stories of decipherment.  An incredible amount of work has been done in computer science with respect to natural language processing in the last 40 years and I am interested in finding out if anyone in the ANE community looking at ancient languages is using any of these techniques.

            Mike Matsko

            On Mar 30, 2014, at 9:19 AM, "Peter T. Daniels" <grammatim@...> wrote:

             
            What exactly are you looking for? She doesn't talk about any actual decipherments, but presents some mathematical algorithms. The standard work has been Maurice Pope's volume from Thames & Hudson (it had different titles in the UK and US editions, and for the second edition too), which is reliable on the whole except that in the 2nd edition a reference is added in a footnote to my article on Hincks's decipherment of Akkadian but clearly he didn't even look at it, since the text continues to ascribe the achievement to Rawlinson. (For the demonstration that Rawlinson contributed almost nothing to the decipherment of cuneiform, see my article in the on-line Encyclopedia Iranica, which is at iranica.org.)

            The journalist Andrew Robinson's Lost Languages is a coffee table book mostly about undeciphered scripts but includes some information on decipherments. (It's considerably better than his earlier History of Writing.) He also wrote little biographies of Michael Ventris and of Thomas Young.

            You'll find lots of references to the primary publications of decipherments in my chapter in The World's Writing Systems (1996), and overviews in the Oxford Encyclopedia of the Archeology of the Near East and in Civilizations of the Ancient Near East (Sasson et al.).. 
            --
            Peter T. Daniels grammatim@...


            From: Michael Matsko <msm@...>
            To: ANE-2@yahoogroups.com
            Sent: Saturday, March 29, 2014 10:37 PM
            Subject: [ANE-2] Archeological Decipherment

             
            Does anyone know of a more recent study of archeological decipherment than E. J. W. Barber's 1974 Archeological Decipherment? I am aware of Kevin Knight's work (at USC's ISI) do not know of much else.

            Mike Matsko
            Arlington, VA
          • Michael Matsko
            No. I think that the consensus on currently undeciphered texts is that if there was a sizable corpus available, then they would have been deciphered. My
            Message 5 of 12 , Mar 30, 2014
            • 0 Attachment
              No.  I think that the consensus on currently undeciphered texts is that if there was a sizable corpus available, then they would have been deciphered.  

              My question was more in line with inquiring if Barber's work has been expanded to include current NLP techniques with a thought to rederiving extant decipherments or possibly using statistical techniques to iron out some of the finer points of current decipherments.

              Mike Matsko
              Arlington, VA

              On Mar 30, 2014, at 1:10 PM, "Peter T. Daniels" <grammatim@...> wrote:

               

              Can you point to a sizable corpus of undeciphered texts where a computational algorithm might be useful? IIRC Barber offered a mathematicization of what Kober and Ventris had done with Linear B. 
              --
              Peter T. Daniels grammatim@...
              Jersey City


              From: Michael Matsko <msm@...>
              To: ANE-2@yahoogroups.com
              Sent: Sunday, March 30, 2014 11:16 AM
              Subject: Re: [ANE-2] Archeological Decipherment

               
              Peter,

              Thank you for your detailed and comprehensive response.

              To your question.  I am looking more for references to algorithmic aids to decipherment than stories of decipherment.  An incredible amount of work has been done in computer science with respect to natural language processing in the last 40 years and I am interested in finding out if anyone in the ANE community looking at ancient languages is using any of these techniques.

              Mike Matsko

              On Mar 30, 2014, at 9:19 AM, "Peter T. Daniels" <grammatim@...> wrote:

               
              What exactly are you looking for? She doesn't talk about any actual decipherments, but presents some mathematical algorithms. The standard work has been Maurice Pope's volume from Thames & Hudson (it had different titles in the UK and US editions, and for the second edition too), which is reliable on the whole except that in the 2nd edition a reference is added in a footnote to my article on Hincks's decipherment of Akkadian but clearly he didn't even look at it, since the text continues to ascribe the achievement to Rawlinson. (For the demonstration that Rawlinson contributed almost nothing to the decipherment of cuneiform, see my article in the on-line Encyclopedia Iranica, which is at iranica.org.)

              The journalist Andrew Robinson's Lost Languages is a coffee table book mostly about undeciphered scripts but includes some information on decipherments. (It's considerably better than his earlier History of Writing.) He also wrote little biographies of Michael Ventris and of Thomas Young.

              You'll find lots of references to the primary publications of decipherments in my chapter in The World's Writing Systems (1996), and overviews in the Oxford Encyclopedia of the Archeology of the Near East and in Civilizations of the Ancient Near East (Sasson et al.).. 
              --
              Peter T. Daniels grammatim@...


              From: Michael Matsko <msm@...>
              To: ANE-2@yahoogroups.com
              Sent: Saturday, March 29, 2014 10:37 PM
              Subject: [ANE-2] Archeological Decipherment

               
              Does anyone know of a more recent study of archeological decipherment than E. J. W. Barber's 1974 Archeological Decipherment? I am aware of Kevin Knight's work (at USC's ISI) do not know of much else.

              Mike Matsko
              Arlington, VA

            • Peter T. Daniels
              I haven t heard of such a thing, but I have no interest in computational linguistics, and the one scholar who has tried to develop a computational theory of
              Message 6 of 12 , Mar 30, 2014
              • 0 Attachment
                I haven't heard of such a thing, but I have no interest in computational linguistics, and the one scholar who has tried to develop a "computational theory of writing systems," Richard Sproat, insists that the Indus Valley script is not writing. 
                --
                Peter T. Daniels grammatim@...
                Jersey City

                From: Michael Matsko <msm@...>
                To: "ANE-2@yahoogroups.com" <ANE-2@yahoogroups.com>
                Sent: Sunday, March 30, 2014 4:00 PM
                Subject: Re: [ANE-2] Archeological Decipherment

                 
                No.  I think that the consensus on currently undeciphered texts is that if there was a sizable corpus available, then they would have been deciphered.  

                My question was more in line with inquiring if Barber's work has been expanded to include current NLP techniques with a thought to rederiving extant decipherments or possibly using statistical techniques to iron out some of the finer points of current decipherments.

                Mike Matsko
                Arlington, VA

                On Mar 30, 2014, at 1:10 PM, "Peter T. Daniels" <grammatim@...> wrote:

                 
                Can you point to a sizable corpus of undeciphered texts where a computational algorithm might be useful? IIRC Barber offered a mathematicization of what Kober and Ventris had done with Linear B. 
                --
                Peter T. Daniels grammatim@...
                Jersey City
              • richardwsproat
                Hi Mike: The most successful work as been that of Ben Snyder, Regina Barzilay and Kevin Knight on Ugaritic. For example: http://people.csail.mit.edu/
                Message 7 of 12 , Mar 31, 2014
                • 0 Attachment
                  Hi Mike:

                  The most "successful" work as been that of Ben Snyder, Regina Barzilay and Kevin Knight on Ugaritic. For example:


                  I put "successful" in quotes because on the one hand they did succeed in (re)deciphering Ugaritic by a technique that finds parallels with a closely related language (Hebrew in this case).  And, predictably, it did generate a lot of press at the time.

                  On the other hand, Ugaritic is of course special in that it's a segmental system, with a small alphabet, that is closely related in structure if not outward form to other writing systems (Semitic abjads).  And there are closely related languages to be found. Obviously neither of those situations holds in most cases of archaeological decipherment.

                  There is of course also the question of how much text is needed to build a reliable computational model, and whether one generally has sufficient text for a given ancient script.

                  Richard Sproat
                  Google, Inc
                • richardwsproat
                  By the way, I am not sure there is a consensus on currently undeciphered texts to the effect that if there was a sizable corpus available, then they would
                  Message 8 of 12 , Mar 31, 2014
                  • 0 Attachment
                    By the way, I am not sure there is "a consensus on currently undeciphered texts" to the effect "that if there was a sizable corpus available, then they would have been deciphered."

                    There certainly are undeciphered systems where the amount of text is reasonable --- say a few thousand glyphs --- but which remain uncracked.  A good example is rongorongo, where the conditions for decipherment could hardly be better: We have several thousand glyphs' worth of long texts; we know the language spoken by the Easter Islanders; and we even know a lot from ethnographic studies about how this system was used. Yet despite attempts dating back to the 19th century, nobody has made any progress (though there have been some interesting structural analyses of the glyphs). One possibility is that this may not have been a script in the normal sense but rather more of a mnemonic system, like the Naxi pictographic "script" (though that also has some linguistic elements).

                    Another example, of course, is the Indus Valley symbols. There the situation is hardly as favorable as in the case of rongorongo: we do not know what language(s) the inhabitants of the IVC spoke' and the texts are all exceedingly short, though there are a lot of them.  Still, despite many attempts, nobody has come up with anything that is convincing to many people besides the decipherer him or herself.  There are people who remain convinced that the underlying language was some form of Dravidian, and thus that Parpola's original work was on the right track: but since that work essentially ground to a halt  over 40 years ago, with but a handful of proposed readings of symbols to its credit, there is no rational basis for that conviction. And as has been pointed out, there are those of us who believe that the IV symbols were not writing, and that therefore it is a waste of time to try to decipher them using linguistic decipherment techniques.

                    Richard Sproat
                    Google, Inc

                  • Peter T. Daniels
                    This unpublished, undated paper, previously unknown to me, begins with so many unwarranted assumptions that it is useless as a guide for computerized
                    Message 9 of 12 , Mar 31, 2014
                    • 0 Attachment
                      This unpublished, undated paper, previously unknown to me, begins with so many unwarranted assumptions that it is useless as a guide for computerized decipherment.

                      1) It assumes that the unknown is in a language closely related to a known _and identified_ language. This is almost never the case in the history of decipherment. Choosing which language(s) to compare is often a huge problem. And what if the first Ugaritic tablets had only been the few written in Hurrian? While everyone thought Northwest Semitic was the likeliest guess given the location, no one expected to find Hurrian texts there.

                      2) It assumes on the basis of the number of characters in the script (30 in this case) that the script is "alphabetic" (in this case, actually, abjadic, though here the distinction isn't important). But that wouldn't have worked for Celtiberian (deciphered in 1955 after centuries of attempts) or for Meroitic (deciphered in 2010 or so): both have about 30 characters, but the former is part syllabic and part segmental, and the latter is abugidic.

                      3) It assumes that the character inventory is readily determined. This has been a huge problem with the Byblos script: analyses have ranged from 22 different characters to 114 different characters.

                      4) It relies for information on the decipherment of Ugaritic on one highly superficial account by a journalist (Andrew Robinson, Lost Languages) and one chapter in a handbook (by someone who was not the right choice to contribute the article). (The authors even get the date of the achievement wrong.) It does not consult either the original publications by the three scholars who, working for the most part independently and with very different methods, achieved three converging decipherments, or the two detailed articles on the topic, by Alan Corré (1966) and by Peggy Day (2002). My own contribution on the topic was presented at the Rutgers NACAL meeting in 2012.

                      I have no competence to evaluate the algorithm(s) presented, but given the tremendous amount of "assuming what is to be proved," i.e. begging the question, it seems unlikely that it will be useful in any other decipherments (such as Linear A, Indus, rongorongo [which is generally recognized as not a writing system], or even, heaven help us, the Ph**st*s D*sk]).
                      --
                      Peter T. Daniels grammatim@...
                      Jersey City

                      From: "rws@..." <rws@...>
                      To: ANE-2@yahoogroups.com
                      Sent: Monday, March 31, 2014 5:30 AM
                      Subject: [ANE-2] Re: Archeological Decipherment

                       
                      Hi Mike:

                      The most "successful" work as been that of Ben Snyder, Regina Barzilay and Kevin Knight on Ugaritic. For example:


                      I put "successful" in quotes because on the one hand they did succeed in (re)deciphering Ugaritic by a technique that finds parallels with a closely related language (Hebrew in this case).  And, predictably, it did generate a lot of press at the time.

                      On the other hand, Ugaritic is of course special in that it's a segmental system, with a small alphabet, that is closely related in structure if not outward form to other writing systems (Semitic abjads).  And there are closely related languages to be found. Obviously neither of those situations holds in most cases of archaeological decipherment.

                      There is of course also the question of how much text is needed to build a reliable computational model, and whether one generally has sufficient text for a given ancient script.

                      Richard Sproat
                      Google, Inc
                    • richardwsproat
                      I just want to address a few of the many incorrect statements and misunderstandings below. First, the paper was indeed published: it was published in the 2010
                      Message 10 of 12 , Mar 31, 2014
                      • 0 Attachment

                        I just want to address a few of the many incorrect statements and misunderstandings below.

                        First, the paper was indeed published: it was published in the 2010 proceedings of the Association for Computational Linguistics. Indeed the name of the file in the link is a hint to that, but in any case a simple search on the title would have revealed the true state of affairs (see e.g. http://dl.acm.org/citation.cfm?id=1858788).

                        Second, yes, I said very clearly that this does make a lot of assumptions that are not very realistic, so I was not holding this up as a solution to the general problem.  I didn't think Mike was asking about that.  

                        Third, I happen to know that Snyder did consult with at least one Ugaritic expert, Wayne Pitard at the University of Illinois. I know this because Snyder visited Illinois while I was still there (that was a few years ago before I left my tenured position there by choice, and before I moved to Google, also by choice).

                        Is rongorongo "generally recognized as not a writing system"? Most of the people still working on it, at least, assume it was a writing system.  Macri (who I guess was working on it at one time) even published a chapter in Daniels & Bright where she claimed that it was some sort of syllabic system if you decompose the glyphs right. Of course she presented no evidence for it, which makes me wonder why that chapter was included, but anyway, there are still so many people that consider it to be writing that any general statement such as the one below seems unwarranted.

                        Actually such techniques do work rather well for the Phaistos Disk: I myself have used a much more simple-minded algorithm to decipher Side A of the Disk as Greek. Of course that was complete nonsense, which was the point I was trying to make in doing it: precisely because the text is so short, you can fit it to all kinds of things. And that of course was Chadwick's point about the Disk too.

                        Linear A is irrelevant for the reasons I gave in a different post on this thread: there the issue is not *decipherment*, but the in principle much harder problem of understanding a completely unknown language without any useful points of reference.  Nobody to my knowledge has proposed a computational algorithm for that.

                        Richard Sproat
                        Google, Inc
                        ---In ANE-2@yahoogroups.com, <grammatim@...> wrote :

                        This unpublished, undated paper, previously unknown to me, begins with so many unwarranted assumptions that it is useless as a guide for computerized decipherment.

                        1) It assumes that the unknown is in a language closely related to a known _and identified_ language. This is almost never the case in the history of decipherment. Choosing which language(s) to compare is often a huge problem. And what if the first Ugaritic tablets had only been the few written in Hurrian? While everyone thought Northwest Semitic was the likeliest guess given the location, no one expected to find Hurrian texts there.

                        2) It assumes on the basis of the number of characters in the script (30 in this case) that the script is "alphabetic" (in this case, actually, abjadic, though here the distinction isn't important). But that wouldn't have worked for Celtiberian (deciphered in 1955 after centuries of attempts) or for Meroitic (deciphered in 2010 or so): both have about 30 characters, but the former is part syllabic and part segmental, and the latter is abugidic.

                        3) It assumes that the character inventory is readily determined. This has been a huge problem with the Byblos script: analyses have ranged from 22 different characters to 114 different characters.

                        4) It relies for information on the decipherment of Ugaritic on one highly superficial account by a journalist (Andrew Robinson, Lost Languages) and one chapter in a handbook (by someone who was not the right choice to contribute the article). (The authors even get the date of the achievement wrong.) It does not consult either the original publications by the three scholars who, working for the most part independently and with very different methods, achieved three converging decipherments, or the two detailed articles on the topic, by Alan Corré (1966) and by Peggy Day (2002). My own contribution on the topic was presented at the Rutgers NACAL meeting in 2012.

                        I have no competence to evaluate the algorithm(s) presented, but given the tremendous amount of "assuming what is to be proved," i.e. begging the question, it seems unlikely that it will be useful in any other decipherments (such as Linear A, Indus, rongorongo [which is generally recognized as not a writing system], or even, heaven help us, the Ph**st*s D*sk]).
                        --
                        Peter T. Daniels grammatim@...
                        Jersey City

                        From: "rws@..." <rws@...>
                        To: ANE-2@yahoogroups.com
                        Sent: Monday, March 31, 2014 5:30 AM
                        Subject: [ANE-2] Re: Archeological Decipherment

                         
                        Hi Mike:

                        The most "successful" work as been that of Ben Snyder, Regina Barzilay and Kevin Knight on Ugaritic. For example:


                        I put "successful" in quotes because on the one hand they did succeed in (re)deciphering Ugaritic by a technique that finds parallels with a closely related language (Hebrew in this case).  And, predictably, it did generate a lot of press at the time.

                        On the other hand, Ugaritic is of course special in that it's a segmental system, with a small alphabet, that is closely related in structure if not outward form to other writing systems (Semitic abjads).  And there are closely related languages to be found. Obviously neither of those situations holds in most cases of archaeological decipherment.

                        There is of course also the question of how much text is needed to build a reliable computational model, and whether one generally has sufficient text for a given ancient script.

                        Richard Sproat
                        Google, Inc
                      • Michael Matsko
                        Richard, Thank you for the reference. While Knight, et al., did pretty much work with the best possible case, it is still impressive that it worked at all.
                        Message 11 of 12 , Mar 31, 2014
                        • 0 Attachment
                          Richard,

                             Thank you for the reference.  While Knight, et al., did pretty much work with the best possible case, it is still impressive that it worked at all.  Yes, my original question was more that with all the progress in NLP in the last 40 years, has anyone applied those techniques to the analysis of ancient languages, not as much is this a solved problem (although that would be nice). 
                             I stand corrected with reference to rongorongo and the Indus Valley script, should it be a language.

                          Mike Matsko
                          Arlington, VA
                        • richardwsproat
                          Hi Mike: Sure, I agree that it is impressive that it worked as well as it did. Unfortunately as with all these things, once one has one success, there are bold
                          Message 12 of 12 , Apr 1, 2014
                          • 0 Attachment
                            Hi Mike:

                            Sure, I agree that it is impressive that it worked as well as it did. Unfortunately as with all these things, once one has one success, there are bold excpectations that the whole problem is going to be solved with just a little more work, despite the obvious indications that Ugaritic is a rather special case.

                            Richard Sproat
                            Google, Inc.
                          Your message has been successfully submitted and would be delivered to recipients shortly.