1 1 2362017 1 Systems for Information Transmission over Networks This
invention is concerned with systems, apparatus and methods for sending information over networks. More particularly, it relates to use of speech or voice patterns to identify intermediate or final destinations for information transmitted over a network, such as an internet or intranet.
The Internet or World Wide Web (WWW) uses a system of named domains to access and route information. Resources on the network are identified by Uniform Resource Locators (URLs) which identify an Internet domain or a more specific resource such as a page of information on a server or a file of digital data. The URL translates, at least in part, to an IP (Internet Protocol) address, which identifies a server at which the resource is available. In IP address at present comprises a four byte number which identifies a server to the Internet hardware so that messages can be properly routed. Often, an IP address is expressed as a set of four decimal numbers each less than 256, for example 56.123.78.2. This number indicates both a network and a host within the network. Routing in IP is based upon a network number part of the destination address.
There are future plans for this arrangement to be modified and/or enhanced to allocate a unique identifying number to every so-called "digital resource". This will avoid potential URL length limits. Digital resources will be identified at every scale, from elements on a web page to servers.
The Internet is one example of a computer network based upon Internet Protocol (TCP/IP - Transmission Control Protocol/Internet Protocol). An internet may also, exist separately from the World Wide Web and, when operated within a single organisation is often referred to as an intranet. In what follows, for convenience, 2 specific reference will be made to the Internet, although the principles and techniques described are not limited to this embodiment of an internet, nor do they specifically require use of the Internet Protocol which, essentially, defines a low level standard for message exchange between hardware devices.
Currently, a typical user accesses the Intemet by means of a home computer. However, other Internet access means are also available, including Internet enabled televisions and mobile telephones, for example, using Wireless Access Protocol (WAP). For convenience and simplicity, the example of access via a computer will be used. Typically such access is via a web browser such as Netscape (Registered Trade Mark) or Internet Explorer (Registered Trade Mark).
Information is communicated and exchanged over the Internet by means of a system of named domains. There are different levels of domain and this is reflected in the domain names which, in general, comprise an alphanumeric string. Thus for example, fredbloggs.co.uk, the suffix "uV names the top level domain designating the United Kingdom. Other top level domains include ".com" for internationally registered names, G4.org" for noncommercial organisations, and ".ltd.uk" for private limited companies in the UK.
The allocation of domain names is overseen by ICANN who appoint Domain Name Registrars. These presently include Nominet in the UK, and in the US Network Solutions. Rules have been drawn up for the allocation of domain names and the resolution of disputes. A domain name is allocated to a company or individual and provides a means by which a computer linked to the Internet can access the named domain. Typically, access is provided to a web site, that is a series of files, typically HTML pages, on a computer connected to the Internet. Domain names also facilitate the transmission and exchange of e-mail messages. An e-mail address will normally include the domain name of a recipient or of an e-mail service provider for the recipient.
3 When an Internet user requests access to a specified web site, for example, fredbloggs.com, the request is sent to the user's Internet Service Provider (ISP) which must then determine an IP address to which to forward the message. Typically, the ISP will communicate over the internet with a Domain Name Server (DNS). This will translate the alphanumeric domain name into an IP address using a domain name database. The ISP will then attach the IP address to the message, together with an appropriate reply address, and forward the message through the Internet. In some cases, however, the ISP will maintain a mirror or partial mirror of the DNS database on a local server. As is known to those skilled in the art, a query requesting translation of a name to an address may employ UDP (user datagram protocol) and/or ICMP (internet control message protocol) rather than TCP.
The use of alphanumeric domain names, which typically appear as words, makes identification of web sites easy. However, growing usage of the Internet is leading to a shortage of appropriate domain names. Furthermore, as access to the Internet moves from mainly computer based web browsers to alternative Internet access means such as mobile communications and consumer devices, there is a need to be able to identify a web site without typing in an alphanumeric string on a keyboard. The present invention aims to alleviate these problems.
In accordance with a first aspect of the invention there is therefore provided a method of sending infon-nation over a network, the network having network access means for a user; the method comprising the steps of- i) accepting, at an audio input of the network access means, voice input from the user; ii) passing a representation of the voice input to a server; iii) identifying a destination on the network corresponding to the voice input; and iv) sending information from the user to the matched destination.
An embodiment of this method allows a user to speak into their computer (or other Internet access means) to identify an information destination (final or intermediate), rather than having to type in a domain name. A representation of the spoken input is passed to a server and a destination corresponding to the voice input is then identified.
4 This allows Internet users to register not only "text" domain names such as "fredbloggs.com", but also a corresponding voice pattern, for example, for "fredbloggs.coni". This is possible because typed and spoken domain names can be resolved to separate addresses. In an embodiment digital signals are produced and sent to a Domain Name Server (DNS) when a user types in a domain name ("texf 'domain name) or speaks a Domain Name ("voice pattern" Domain Name) into his computer, telephone or other Internet access device. These are hashed by the Domain Name Server (DNS) using an algorithm to achieve separate and unique values. These are then resolved by the Domain Name Server (DNS) into separate IP addresses/numbers and thus a user's request can be sent to different web sites/servers/network computers.
In one embodiment a user employs an access device, such as a web browser, and either types in a "text" domain name request (e.g. "fredbloggs.com"), which is resolved by a Domain Name Server (DNS) to address or resource "A", or the user may speak a "voice pattern" domain name (e.g. fredbloggs. com") into his/her computer microphone, telephone or other device, the Domain Name Server (DNS) then resolving that request to address or resource "B". Thus, in this embodiment the number of domain names available for registration can effectively be doubled by allowing registration of either or both of the known type of "text" domain names and one or more "voice pattern" domain names. Alternatively, the two (or more) requests can map to the same address/resource.
The system also allows users to register "voice pattern" domain names which translate directly to an IP address (rather than being associated with a particular text domain name). This provides a qualitatively different way of providing a link between a web site and an identification of it which is symbolically meaningful to the user (i.e. a spoken rather than a typed name) and which can be translated to an IP address. The present usage by the Internet, as it stands, of only four bytes for an IP address imposes a limitation on the potential number of domains, but this is not a limitation of the invention.
In a preferred embodiment, the network is an Internet Protocol Network and a destination comprises an Internet protocol address. The information sent may include an Internet address of the user as a return address. The destination may comprise a network domain or may be resolved to a digital resource such as a specific web page or file. The server to which the representation of the voice input is passed is preferably a domain name server or a domain resource server. The destination may additionally or alternatively identify part or whole of an e-mail address. Preferably, the step of identifying a destination is performed by the server.
In one embodiment, a digitised version of the speech or voice input is mathematically transformed before being used to identify a network destination. The transformation may take the form of a "hash table" type transform. This transform is known to those skilled in the art to convert voice input or speech pattern to a hash value. The hash value comprises one or more numbers uniquely associated with a particular speech pattern. The hash value is then used to identify the network destination. Hashing can take place at the user or DNS server end or at an intermediate point. Alternatively, the captured, digitised voice input may be passed to the server without such transformation and the captured input is then used to either generate a unique target destination or to generate a list of potential targets from which a best match destination can be selected.
In another embodiment, the step of identifying matches of voice input to a spoken representation of a network domain name. Thus, for example, a spoken message containing "fredbloggs" would be matched to one or more domains including the words "fredbloggs", for example fredbloggs.com. The system can then be configured to select as the destination either the domain whose name matches the spoken input or a different domain. This latter alternative allows voice-based -Internet access to be specifically directed to a destination which is adapted for voice rather than textbased use.
Other aspects of the invention are set out in independent claims 13, 18, 20 and 2 1, with their preferred features in the independent claims 6 The above and other aspects of the invention will now be finiher described, by way of example only, with reference to the accompanying figures in which:
Figure 1 shows an Internet; and Figure 2 shows data flow between elements of a system embodying the present invention.
In Figure 1, an Internet 10 comprises a TCP/IP computer network or web 12 including standard network components such as hubs, routers, switches and the like. Internet service provider 14 is coupled to web 12, for example via a megastream, and PCs 18a and 18b are coupled to ISP 14, typically via an analogue phone line and modem, to allow user access to the Internet. The second ISP shown at 16, is coupled to a mobile telephone base station 20, for Internet access ftom mobile phones 22a and 22b using WAP (Wireless Application Protocol). Also, coupled to web 12 and comprising part of the Internet are domain name servers 24 and 26 and domain resource server 28. A domain "fredbloggs.com" is provided by first server 30 and a second domain "fred2.com" is provided by a second server 32. Servers 30 and 32 provide web pages accessible to Internet access means 18a, 18b, 22a and 22b.
Referring now to Figure 2, this shows the data flow between components of the Internet of Figure 1. In a conventional Internet, when a user wishes to access a web page, the user types a domain name for the web page into a web browser running on a PC. The web browser passes the request to an ISP which forwards it to a domain name server where it is resolved to a unique IP address. The ISP then forwards a request for the user to connect to the identified server, and the server then returns the requested information to the user's IP address. In Figure 2, the user requests access to a web site or enters an e-mail address verbally by speaking the appropriate domain name/URL/address into an Internet access device such as a mobile telephone, or as illustrated, a microphone of computer 1 Sa. The voice pattern is then digitised and transmitted either as a data stream 7 or, preferably, as a file by stub resolver software 40, in a query 42, to domain name server 24.
DNS 24 has access to a database 44 of speech patterns for registered (verbal) domain names and temporarily stores query 42 in a cache 46 whilst database 44 is used to resolve the query to an IP address. If the speech pattern is not available in database 44, domain name server 24 preferably tries to match query 42 in another database accessible via the Intemet or, alternatively, forwards the query to an alternative "foreign" domain name server 48 which responds to its client, DNS 24. Once the IP address has been resolved, a response 50 is returned to stub resolver 40; alternatively an error or exception message may be returned if it was not possible to resolve the IP address.
The voice or speech pattern attached to query 42 can be matched to the database by conventional methods well known to those skilled in the art. For example, a set of voice parameters can be determined and the distance to the nearest matching speech pattern in parameter space can be determined. In a preferred embodiment, the main name server 24 hashes the user's voice pattern to create a single value representative of the pattern. Suitable hash transform algorithms are well known to those skilled in the art. The domain name server then either directly maps the hash value to an IP address or compares the hash value with corresponding numbers for registered speech patterns in database 44 to identify a specific Internet address. If there is no exact match the closest match can be determined or alternatively a list of potential matches can be returned with response 50.
In another embodiment, the speech pattern is hashed at the user end, in the user's Internet access device, for example, PC 18a. The hashing algorithm can, if desired, be performed by dedicated hardware such as an ASIC in mobile phone 22a. In this case, the query need only contain the single value resulting from the processed (hash) speech pattern, which reduces Internet traffic. The domain name server matches the hash value attached to query 42 as described above.
8 The database 44 can readily be constructed and managed in a broadly corresponding manner to that in which conventional domain name databases are constructed and managed. Thus one or more exemplary speech patterns can be obtained on registration of a domain. The stored voice pattern or patterns can either be linked to a separately registered alphanumeric domain name such as "fredbloggs.com", or can provide an entirely separate system of domain registration and identification.
In an embodiment where a voice pattern is a spoken representation of an alphanumeric type registered domain name, there is no need to request an exemplary speech pattern as voice recognition technique can be used to match a spoken domain name to its text equivalent.
No specific language restrictions need be made although if, as it will often be the case, a user's country or region is known, this information can be used to increase the probability of a successful match to an IP address by the domain name server.
At present, an 'http' URL is composed of a domain name which through the present Internet domain resolution system resolves to an IP address of a connected Internet device. Additionally, the URL will often include short or extended 'path' to a specific computer file, which may in practice be a web page, an image file, or program/script file. These and other end points of navigation may be generally referred to as 'digital resources'. Future Internet developments are likely to provide alternative URL protocols to the presently preferred 'http' URL protocol. These include the CNRI 'Handle' System, which plans to identify each and every individual 'digital resource' in a specific and simplified manner, associated servers providing resolution and navigation to the associated Internet device and resource. The applicability of the voice assisted system described herein should be recognised as not being limited to the present 'http' arrangement but is applicable to, and can extend, systems based on alternative arrangements.
9 In some embodiments the registered voice pattern can be resolved not only to a domain and its IP address, but additionally to the equivalent of a full 'http' URL, either by registration of the pattern as such, or by using voice recognition techniques. In this way the system can enable direct voice assisted Internet navigation, not just to a domain or "home page" by virtue of its IP address but to a specific web page or other 'digital resource'. This corresponds to a user typing a complete alphanumeric URL into an Internet browser, requesting to be taken directly to a specified path and file within a domain.
Such an arrangement is particularly advantageous with mobile Internet access devices such as mobile telephones/organisers, where a lengthy typed URL is impractical with WAP browsers. However, "absolute" navigation to WAP pages within domains can be achieved more practically by resolving a voice pattern to a full URL (or its equivalent such as the 'Handle' system developed by the CNRI) rather than simply to an IP address of a registered Domain.
E-mail addresses can be resolved in the same way as web site and domain name addresses. Although the system has been described in the context of domain names, it is equally applicable to resolve spoken requests to digital addresses for digital resources using, for example, a domain resource server.
No doubt many other effective embodiments will occur to those skilled in the art and it should be understood that the invention is not limited to the described embodiments.